WHOLE-CABIN BOARDING SIMULATION
DETERMINISTIC STRESS DATASETS FOR MULTI-REGION DEPENDENCY EVALUATION
Version: Stress Dataset Research Log v1

Purpose
-------
This document begins the final controlled validation programme for the completed
Reactive Dependency-Region Management Architecture.

The preceding architecture report is now frozen. Its final implementation
provides reactive congestion detection, persistent dependency-region lifecycle
tracking, three-tick confirmation, state-change-aware evaluation and hierarchical
dependency-region prioritisation.

The present programme does not redesign that architecture. It investigates
whether deterministic input datasets can be constructed to produce simultaneous
confirmed dependency regions and thereby exercise architectural capabilities that
rarely appeared in the conventional benchmark scenarios.

The architecture is therefore the fixed reference implementation. Dataset
construction becomes the independent variable.


======================================================================
RESEARCH BOUNDARY
======================================================================

This research does not reopen or replace the conclusions of the completed
Reactive Dependency-Region Management Architecture.

It introduces one new independent research variable:

    DETERMINISTIC STRESS DATASET CONSTRUCTION

The programme may alter deterministic passenger manifests, seat-distribution
patterns and queue ordering only when those changes are explicitly documented as
properties of the dataset under test.

It does not introduce:

- probabilistic human behaviour;
- passenger hesitation or personal judgement;
- asynchronous tick execution;
- overtaking;
- reverse aisle movement;
- aisle switching;
- movement beyond an assigned row;
- seat reassignment;
- predictive congestion management;
- advance aisle-space reservation;
- airline policy unless later isolated as a separate deterministic dataset
  variable.

The completed architecture must process each dataset without being changed to
accommodate the expected result.


======================================================================
FIXED REFERENCE ARCHITECTURE
======================================================================

The following architectural components remain enabled and unchanged throughout
this research programme:

- synchronous global movement;
- maximum one forward aisle tile per passenger per tick;
- one unique assigned seat per passenger;
- fixed serving aisle for every passenger;
- no movement beyond the assigned row;
- deterministic stand-yield-sit-reseat transactions;
- middle-bank reservation;
- dependency-aware precedence;
- adaptive rear-boundary forward yield;
- guarded forward-yield checks;
- row-bounded congestion clustering;
- persistent dependency-region lifecycle tracking;
- three-tick region confirmation;
- state-change-gated evaluation;
- deterministic hierarchical region prioritisation;
- 1,500-tick no-seat-progress stall detector;
- 30,000 maximum ticks.

Any later change to one of these components would constitute a new architecture
programme and must not be presented as a stress-dataset experiment.


======================================================================
CONTROLLED CONDITIONS
======================================================================

Within every paired experiment, the following must remain identical unless the
experiment explicitly identifies one of them as the dataset variable:

- cabin configuration;
- total seat capacity;
- occupancy rate;
- scenario seed;
- passenger count;
- entry-headway range;
- synchronous execution;
- passenger movement rules;
- seat-event prerequisites;
- blocker movement;
- yield-space rules;
- middle-bank reservation;
- dependency-region detection;
- lifecycle matching, updating and dissolution;
- confirmation threshold;
- state-change signature;
- region-priority hierarchy;
- stall and maximum-tick limits;
- output instrumentation.

Experiment 1 changes the deterministic dataset family while preserving the
entire architecture and the scenario-level cabin, occupancy, headway and seed.


======================================================================
DATASET CONSTRUCTION CONTRACT
======================================================================

A deterministic stress dataset is a reproducible passenger manifest and entry
ordering designed to increase the opportunity for demanding dependency-region
structures without modifying passenger movement after admission.

All stress datasets must obey the following rules:

1. Every passenger retains one unique assigned seat.
2. Every passenger retains one fixed serving aisle.
3. Dataset construction completes before simulation movement begins.
4. Dataset generation is reproducible from the recorded seed.
5. The boarding engine receives a completed manifest and queue; it does not know
   why a passenger was selected or ordered.
6. No stress generator may create two passengers for the same seat.
7. No generator may place a passenger directly into the aisle or a seat-event
   state.
8. No generator may alter movement speed, event duration or blocker rules.
9. No generator may guarantee or fabricate a dependency region.
10. Failure to create simultaneous regions is a valid dataset result and must be
    documented honestly.


======================================================================
NEW TERMINOLOGY
======================================================================

Deterministic Stress Dataset
----------------------------
A reproducible passenger manifest and entry-order dataset deliberately designed
to expose demanding but valid congestion structures to the frozen architecture.

Standard Deterministic Dataset
------------------------------
The established reference generator using seeded random seat selection and
independently shuffled aisle queues.

Stress Dataset Family
---------------------
A named group of deterministic datasets sharing the same construction principle,
such as row-zone concentration, blocker-prone seating or alternating aisle
admission.

Dataset Characteristic
----------------------
A measurable property of an input dataset, such as row-zone concentration,
blocker-prone seat share or left/right aisle balance.

Confirmed Region Concurrency
----------------------------
The number of dependency regions that have simultaneously survived the
three-tick confirmation threshold.

Region Competition
------------------
A tick on which more than one confirmed, state-change-eligible dependency region
requires hierarchical evaluation.

Competition Window
------------------
A continuous tick interval during which two or more confirmed dependency regions
coexist.

Priority Utilisation
--------------------
Observed use of the completed hierarchical prioritisation layer to rank genuine
simultaneous confirmed regions.

Negative Dataset Finding
------------------------
A valid result in which a stress dataset does not produce the intended region
competition. Such a result informs later dataset refinement without justifying
an architecture change.


======================================================================
RESEARCH OBJECTIVES
======================================================================

1. Construct reproducible deterministic datasets capable of producing multiple
   dependency regions within the same simulation.
2. Progress from detected-region concurrency to simultaneous confirmed-region
   concurrency.
3. Produce genuine priority competitions without modifying the completed
   architecture.
4. Identify which input characteristics increase or reduce multi-region
   dependency formation.
5. Separate dataset effects from entry headway and all movement rules.
6. Retain negative findings when apparently stressful datasets simplify rather
   than intensify congestion.
7. Build a reusable catalogue of deterministic benchmark datasets for future
   validation.
8. Complete a final controlled test of the frozen architecture.


======================================================================
EXPERIMENT 1 - BASELINE DETERMINISTIC STRESS DATASET CONSTRUCTION
======================================================================

Connection to Amit Amlani's Original People-Movement Algorithm
--------------------------------------------------------------
Amit Amlani's original movement algorithm operates on an already constructed
one-dimensional occupancy dataset. The completed boarding architecture similarly
processes the manifest and aisle queues that it receives.

Experiment 1 returns to this foundational distinction. It does not change the
movement algorithm. It changes only the deterministic data supplied to that
algorithm and observes whether the same fixed architecture exposes different
congestion structures.

Purpose
-------
Establish a reusable deterministic stress-dataset generator and compare its
first conservative stress family against the standard deterministic dataset.

The first experiment is primarily a foundation and calibration experiment. It
is not required to produce simultaneous confirmed dependency regions. Its role
is to verify that:

- dataset construction is reproducible;
- the frozen architecture remains unchanged;
- paired datasets retain the same scenario-level cabin, occupancy, headway and
  seed;
- dataset characteristics are recorded explicitly;
- region-concurrency and priority evidence can be compared cleanly.

Paired Modes
------------
Each of the 30 scenarios is executed twice.

1. STANDARD DETERMINISTIC DATASET

   The reference generator selects occupied seats from a seeded shuffled seat
   catalogue and independently shuffles the left- and right-aisle queues.

2. BASELINE DETERMINISTIC STRESS DATASET

   The frozen architecture receives a reproducible composite stress manifest.
   Passenger seats are preferentially selected from two separated row zones,
   blocker-prone seats receive greater selection weight and each serving-aisle
   queue is ordered deterministically around the same two zones.

The second mode changes input data only. The architecture, movement contract,
confirmation threshold, state-change gate and priority hierarchy are identical.

Independent Variable
--------------------

    DATASET GENERATION FAMILY

- standard seeded manifest and shuffled queues; versus
- baseline deterministic two-zone stress manifest and stress-ordered queues.

Baseline Stress Family Design
-----------------------------
The first stress family uses three documented dataset characteristics:

1. TWO SEPARATED ROW ZONES

   Seats near approximately one-third and two-thirds of cabin depth receive
   greater deterministic selection weight. The separation is intended to create
   more than one local area of potential congestion rather than one continuous
   cabin-wide concentration.

2. BLOCKER-PRONE SEAT EMPHASIS

   Window and deeper seats receive greater weight than aisle-adjacent seats. This
   increases the potential for valid seated blockers without changing any seat
   event after the manifest is created.

3. DETERMINISTIC ZONE-ORIENTED QUEUE ORDER

   Each aisle queue is ordered reproducibly so passengers associated with the
   two stress zones enter under the existing headway mechanism. The generator
   does not bypass headway and does not insert passengers directly into the
   aisle.

This is intentionally a composite baseline family. Later experiments may isolate
or ablate individual characteristics after the generator and evidence model have
been validated.

Controlled Conditions
---------------------

- identical cabin configuration within each pair;
- identical occupancy rate;
- identical passenger count;
- identical entry-headway range;
- identical scenario seed;
- identical frozen architecture;
- identical synchronous tick execution;
- identical movement constraints;
- identical stall detector and maximum ticks;
- no human behaviour;
- no asynchronous execution;
- no reverse movement, overtaking, aisle switching or assigned-row overshoot.

Dataset Evidence Collected
--------------------------

- dataset family;
- row-zone concentration percentage;
- blocker-prone seat share;
- left/right aisle-assignment balance;
- passenger count and occupancy;
- entry-headway range;
- dependency regions detected;
- regions created, updated, confirmed and dissolved;
- maximum simultaneous active regions;
- regions active at termination;
- state-change evaluations suppressed and reopened;
- priority competitions;
- lower-priority deferrals;
- cluster selections and cluster-started events;
- complete cabins and passengers seated;
- simulation ticks;
- passengers remaining outside;
- residual stall family.

Experimental Hypothesis
-----------------------
A deterministic two-zone, blocker-emphasised dataset may increase spatially
separated dependency activity relative to the standard dataset. However, the
same ordering may also simplify flow by making local seat events complete more
coherently. Either result is scientifically valid.

The primary Experiment 1 question is therefore not whether the stress dataset
improves or worsens boarding. It is whether the dataset layer is reproducible,
measurable and capable of changing dependency-region evidence without changing
the architecture.

Interpretation Safeguards
-------------------------

- A faster stress run is not automatically a failed experiment.
- A slower stress run is not automatically evidence of better stress.
- More detected regions do not necessarily mean more confirmed concurrent
  regions.
- Multiple regions created at different times do not constitute simultaneous
  dependency competition.
- Priority utilisation requires genuine simultaneous confirmed eligible regions.
- The architecture must not be changed merely because the first stress family
  fails to create competition.
- Any improvement or regression in boarding outcome must be reported as a
  dataset effect.
- Composite stress characteristics must be separated in later experiments before
  causal claims are made about a single characteristic.

Primary Evaluation Order
------------------------

1. Correct experiment identity and deterministic replay.
2. Dataset-characteristic differences between paired modes.
3. Maximum simultaneous active regions.
4. Confirmed-region concurrency.
5. Genuine priority competitions and deferrals.
6. Regions created, confirmed and dissolved.
7. Cluster selections and cluster-started events.
8. Complete cabins and passengers seated.
9. Residual stall classifications.
10. Simulation ticks as a secondary measure.

Expected Outcomes
-----------------

Outcome A - Simultaneous confirmed regions appear

The baseline stress family immediately exercises hierarchical prioritisation and
provides a foundation for controlled refinement.

Outcome B - More regions appear, but not simultaneously

The generator increases congestion activity but requires stronger temporal
alignment in a later dataset.

Outcome C - Fewer regions appear

The deterministic zone ordering simplifies cabin flow. This becomes a useful
negative finding and suggests later experiments should separate seat selection
from queue ordering.

Outcome D - No meaningful difference appears

The baseline generator is too conservative. Later experiments may increase one
dataset characteristic at a time while retaining the same frozen architecture.

Diagram Plan
------------
A diagram should be considered only after results are available. A valid diagram
may compare:

- the standard dispersed manifest;
- the two separated stress zones;
- blocker-prone seat emphasis;
- left/right deterministic queues;
- observed region timelines.

The diagram must distinguish input construction from passenger movement and must
not imply that dependency regions were inserted directly into the simulation.


======================================================================
RUNNING CONCLUSIONS
======================================================================

1. The Reactive Dependency-Region Management Architecture is frozen.
2. Dataset construction is now the independent research variable.
3. Experiment 1 establishes the reusable paired stress-dataset framework.
4. The first stress family combines two separated row zones, blocker-prone seat
   emphasis and deterministic zone-oriented queue ordering.
5. Simultaneous dependency competition must arise from valid deterministic
   simulation state; it must not be manufactured by changing the architecture.
6. Human behaviour, asynchronous execution, backward movement, overtaking,
   aisle switching and assigned-row overshoot remain excluded.
7. Negative dataset findings are retained as evidence for later refinement.
8. Future experiments should isolate individual stress characteristics after the
   baseline generator has been validated.



======================================================================
EXPERIMENT 1 FINDINGS - BASELINE DETERMINISTIC STRESS DATASET CONSTRUCTION
======================================================================

Execution Summary
-----------------
Experiment 1 completed 30 deterministic paired scenarios (60 executions) using
the same fixed architecture, cabin characteristics, occupancy, headway and
scenario seeds in each pair.

The standard dataset and baseline composite stress dataset each completed 27 of
30 cabins. However, the internal evidence differed substantially.

Overall Paired Result
---------------------

- Standard-dataset average completion: 99.71%.
- Composite stress-dataset average completion: 96.82%.
- Standard complete cabins: 27 of 30.
- Composite stress complete cabins: 27 of 30.
- Improved stress pairs: 3.
- Unchanged pairs: 24.
- Worse stress pairs: 3.
- Standard total passengers seated: 7,649.
- Composite stress total passengers seated: 7,391.
- Net seated-passenger difference: -258.
- Standard dependency regions detected: 37.
- Composite stress dependency regions detected: 13.
- Standard regions reaching confirmation: 22.
- Composite stress regions reaching confirmation: 9.
- Standard cluster selections: 25.
- Composite stress cluster selections: 6.
- Standard cluster-started events: 31.
- Composite stress cluster-started events: 15.
- Genuine priority competitions: 0 in both modes.

Principal Findings
------------------

1. The dataset-generation layer was successfully separated from the architecture.

Both paired modes used the frozen Reactive Dependency-Region Management
Architecture. The experiment therefore confirmed that manifests and aisle queues
can be varied reproducibly without changing movement, clustering, lifecycle,
confirmation, state-change gating or prioritisation.

2. The composite stress family did not create multi-region competition.

No simultaneous confirmed-region priority competition occurred. The dataset
therefore did not yet exercise the hierarchical prioritisation layer.

3. The composite stress family reduced dependency-region activity.

The stress mode produced 13 detected regions compared with 37 under the standard
dataset, and only 9 confirmed regions compared with 22. Cluster selections and
cluster-started events were also lower.

4. Deterministic concentration can simplify flow rather than intensify it.

The combination of two row zones, blocker-prone seat weighting and zone-oriented
queue ordering appears to have created more coherent local seating sequences in
many scenarios. This reduced the fragmented dependency activity that the stress
family was intended to increase.

5. Boarding outcome alone is not a sufficient stress measure.

Both modes completed 27 cabins, yet the stress mode seated 258 fewer passengers
across the three incomplete cases. Conversely, three standard stalls converted
to complete stress runs. The mixed result confirms that completion must be read
alongside region creation, confirmation and competition evidence.

6. The negative result provides a clear experimental refinement.

Because Experiment 1 changed row concentration, blocker-prone seat share and
queue ordering together, no single characteristic can be identified as the
cause of reduced region activity. The next experiment should isolate row-zone
concentration while returning seat-depth weighting and queue ordering to the
standard deterministic method.

Experiment 1 Conclusion
-----------------------
The reusable deterministic stress-dataset framework is validated. The first
composite family did not produce simultaneous dependency competition and often
simplified congestion structure. This is retained as a valid negative dataset
finding. Experiment 2 will isolate row-zone concentration as the sole dataset
variable.


======================================================================
EXPERIMENT 2 - CONTROLLED ROW-ZONE CONCENTRATION
======================================================================

Connection to Amit Amlani's Original People-Movement Algorithm
--------------------------------------------------------------
Amit Amlani's original movement algorithm processes the occupancy dataset it is
given without changing the rules of movement in response to how that dataset was
constructed. Experiment 2 follows the same separation by changing only where
assigned seats are concentrated before movement begins.

Purpose
-------
Determine whether two separated row zones, by themselves, increase spatially
separated dependency activity when blocker-prone seat weighting and specialised
queue ordering are removed.

Paired Modes
------------
Each of the 30 scenarios is executed twice using identical cabin configuration,
occupancy, passenger count, headway, scenario seed and frozen architecture.

1. STANDARD DETERMINISTIC DATASET

   Occupied seats are selected from a seeded shuffled seat catalogue. Left- and
   right-aisle queues are independently shuffled.

2. CONTROLLED ROW-ZONE CONCENTRATION DATASET

   Occupied seats are selected preferentially from two separated row zones near
   one-third and two-thirds of cabin depth. All seat positions inside those zones
   receive equal treatment; window and deep seats receive no additional weight.
   Left- and right-aisle queues are independently shuffled exactly as in the
   standard mode.

Independent Variable
--------------------

    ROW-ZONE CONCENTRATION

- dispersed seeded seat selection; versus
- deterministic concentration around two separated row zones.

Removed Experiment 1 Composite Variables
----------------------------------------

The following Experiment 1 characteristics are deliberately removed:

- blocker-prone seat emphasis;
- zone-oriented queue ordering;
- blocker-first queue tie-breaking.

This permits a narrower causal interpretation of any change in region evidence.

Controlled Conditions
---------------------

- frozen Experiment 23 architecture;
- identical cabin, occupancy and passenger count within each pair;
- identical entry-headway range;
- identical scenario seed;
- independently shuffled aisle queues in both modes;
- identical seat-event and blocker rules;
- synchronous execution;
- no backward movement, overtaking, aisle switching or assigned-row overshoot;
- no human behaviour or asynchronous execution.

Evidence Collected
------------------

- row-zone concentration percentage;
- blocker-prone seat share as a diagnostic control;
- aisle-assignment balance;
- dependency regions detected;
- regions created, confirmed and dissolved;
- maximum simultaneous active regions;
- confirmed-region concurrency;
- priority competitions and lower-priority deferrals;
- cluster selections and starts;
- complete cabins and passengers seated;
- residual stall family;
- simulation ticks as a secondary measure.

Experimental Hypothesis
-----------------------
If separated spatial concentration is sufficient to create independent local
congestion, the row-concentrated dataset should increase concurrent region
activity despite retaining standard shuffled queue order. If region activity
remains lower, row concentration itself may promote coherent completion rather
than competition.

Interpretation Safeguards
-------------------------

- Two row zones do not constitute two dependency regions.
- Regions created at different times do not constitute concurrency.
- A priority competition requires simultaneous confirmed and eligible regions.
- Changes in blocker-prone seat share must be reported, even though no explicit
  blocker weighting is used.
- The architecture must not be changed to force the expected result.
- A negative result will motivate a later isolated seat-depth or admission-order
  experiment rather than reopening the architecture.

Primary Evaluation Order
------------------------

1. Dataset identity and deterministic execution.
2. Row-zone concentration difference.
3. Maximum simultaneous active and confirmed regions.
4. Genuine priority competitions.
5. Regions detected and confirmed.
6. Cluster selections and starts.
7. Complete cabins and seated totals.
8. Residual stall families.
9. Simulation ticks as a secondary measure.

Diagram Plan
------------
A later diagram may compare dispersed seat selection with two separated row
concentration zones while showing that both modes retain independently shuffled
left- and right-aisle queues. It must not depict dependency regions as inserted
or guaranteed by the dataset generator.


======================================================================
RUNNING CONCLUSIONS AFTER EXPERIMENT 1
======================================================================

1. The frozen architecture remains the fixed reference implementation.
2. Experiment 1 validated the reusable deterministic dataset layer.
3. The first composite stress family reduced rather than increased region
   activity and produced no priority competition.
4. Composite dataset variables must now be separated before causal claims are
   made.
5. Experiment 2 isolates row-zone concentration and restores standard shuffled
   queue ordering and neutral seat-depth selection.
6. Simultaneous dependencies must emerge from valid deterministic state.
7. Human behaviour, asynchronous execution and movement-rule changes remain
   excluded.

======================================================================
END OF DETERMINISTIC STRESS DATASET RESEARCH LOG v2 - EXPERIMENT 2 DESIGN
======================================================================

======================================================================
EXPERIMENT 2 FINDINGS - CONTROLLED ROW-ZONE CONCENTRATION
======================================================================

Purpose Reviewed
----------------
Experiment 2 isolated deterministic row-zone concentration from the composite
stress variables introduced in Experiment 1. Both paired modes retained the
frozen Reactive Dependency-Region Management Architecture, independently
shuffled aisle queues, neutral seat-depth selection and identical movement
rules. Only the spatial distribution of occupied rows changed.

Overall Paired Result
---------------------

- Scenario pairs: 30.
- Scenario executions: 60.
- Standard average completion: 99.71%.
- Row-concentration average completion: 98.92%.
- Improved row-concentration pairs: 2.
- Unchanged pairs: 22.
- Worse row-concentration pairs: 6.
- Net seated-passenger difference: -75.
- Standard complete cabins: 27 of 30.
- Row-concentration complete cabins: 24 of 30.
- Standard stalls converted to complete cabins: 2.
- Standard dependency regions: 37.
- Row-concentration dependency regions: 36.
- Standard confirmed regions: 22.
- Row-concentration confirmed regions: 13.
- Standard priority competitions: 0.
- Row-concentration priority competitions: 7.
- Maximum simultaneous active regions: 2 in both modes.

Principal Findings
------------------

1. Row-zone concentration changed congestion geometry without consistently
   increasing overall dependency-region volume.

The row-concentrated datasets produced almost the same total number of detected
regions as the standard datasets, but fewer regions reached confirmation. This
shows that spatial concentration can change the persistence and timing of
congestion even when total region counts remain similar.

2. The isolated variable produced both beneficial and harmful outcomes.

Two standard stalls converted to complete cabins, while six paired scenarios
seated fewer passengers under row concentration. The result therefore does not
support row concentration as a generally beneficial or generally harmful
boarding arrangement.

3. Genuine simultaneous-region competition was observed.

Unlike the previous benchmark and Experiment 1, the row-concentration datasets
produced seven priority competitions. This is the first direct evidence within
the stress-dataset programme that deterministic manifest construction alone can
exercise the hierarchical prioritisation layer of the frozen architecture.

4. Concurrency did not require human behaviour or asynchronous ticks.

The priority competitions emerged under synchronous deterministic movement,
fixed headway rules and unchanged architecture. Dataset geometry alone was
sufficient to create overlapping confirmed-region eligibility.

5. Row concentration is therefore a useful stress component, but not a complete
   causal explanation.

Because row concentration also changed which seats happened to be selected, the
next experiment must isolate seat-depth distribution directly while restoring
dispersed row selection. This will determine whether blocker-prone seats increase
confirmed-region activity independently of row concentration.

Experiment 2 Conclusion
-----------------------
Controlled row-zone concentration is supported as a valid deterministic stress
variable. It generated the first seven genuine priority competitions without
changing the frozen architecture, human behaviour or synchronous execution.
However, its mixed completion effects and reduced confirmation count show that
it should be treated as one stress-building component rather than a complete
solution for repeatable multi-region generation.


======================================================================
EXPERIMENT 3 - CONTROLLED BLOCKER-PRONE SEAT CONCENTRATION
======================================================================

Connection to the Completed Architecture
----------------------------------------
The Reactive Dependency-Region Management Architecture remains frozen. This
experiment changes only the manifest's seat-depth distribution so that the
existing blocker, yield-space and seat-event rules are exercised more often.
The architecture is not informed that a stress dataset is in use and receives
only the completed deterministic passenger manifest.

Purpose
-------
Determine whether deterministic concentration of window and deep inner seats
increases persistent blocker-dependent seat events, confirmed dependency regions
and simultaneous priority competitions when row selection and aisle queues
remain dispersed.

Controlled-Variable Clarification
---------------------------------
Increasing the probability of blocker-prone assigned seats does not alter a
controlled movement variable. The following remain controlled and unchanged:

- the definition of a blocker;
- the stand-yield-sit-reseat transaction;
- the number of yield tiles required by a given seat event;
- blocker movement direction;
- synchronous tick execution;
- passenger movement speed;
- entry-headway range;
- clustering, lifecycle, confirmation, gating and prioritisation rules.

The assigned-seat distribution is intentionally the independent dataset
variable in Experiment 3. Therefore individual seat assignments cannot remain
identical between the two paired modes; instead, cabin type, occupancy,
passenger count, headway, scenario seed and all architecture and movement rules
remain controlled. This distinction is essential to the new research programme,
where dataset construction is the subject under investigation.

Paired Modes
------------

1. STANDARD DETERMINISTIC DATASET

   Occupied seats are selected from a seeded shuffled catalogue. Left- and
   right-aisle queues are independently shuffled.

2. CONTROLLED BLOCKER-PRONE SEAT CONCENTRATION DATASET

   Occupied seats are selected preferentially by deterministic blocker potential.
   Window seats and deep inner seats are selected before easier aisle-side seats.
   Rows remain dispersed through seeded tie-breaking, and both aisle queues are
   independently shuffled exactly as in the standard mode.

Independent Variable
--------------------

    ASSIGNED-SEAT DEPTH DISTRIBUTION

- seeded neutral seat selection; versus
- deterministic concentration of blocker-prone window and deep inner seats.

Controlled Conditions
---------------------

- frozen Experiment 23 architecture;
- identical cabin configuration within each pair;
- identical occupancy rate and passenger count;
- identical entry-headway range;
- identical scenario seed;
- dispersed row selection in both modes;
- independently shuffled aisle queues in both modes;
- unchanged seat-event, blocker and yield-space rules;
- synchronous execution;
- no backward movement, overtaking, aisle switching or assigned-row overshoot;
- no human behaviour or asynchronous execution.

Evidence Collected
------------------

- blocker-prone seat share;
- row-zone concentration as a diagnostic control;
- aisle-assignment balance;
- single- and multiple-blocker seat events;
- temporary aisle reoccupations;
- dependency regions detected;
- regions created, confirmed and dissolved;
- maximum simultaneous active regions;
- priority competitions and lower-priority deferrals;
- cluster selections and starts;
- complete cabins and passengers seated;
- residual stall family;
- simulation ticks as a secondary measure.

Experimental Hypothesis
-----------------------
If blocker-prone seat depth is an important contributor to independent congestion
zones, the concentrated dataset should increase multiple-blocker events,
confirmed dependency regions or priority competitions while row concentration
remains near the standard distribution. A null result would indicate that seat
depth alone is insufficient and that later experiments should combine it with
spatial or admission-pattern stress.

Interpretation Safeguards
-------------------------

- More window seats do not automatically constitute more dependency regions.
- More blocker events do not automatically constitute simultaneous regions.
- A priority competition requires simultaneous confirmed and eligible regions.
- Row-zone concentration must remain a diagnostic control and be reported.
- Different seat assignments are the intended independent-variable change, not
  a breach of the movement contract.
- The architecture must not be modified to force the expected outcome.
- Any worse completion outcome must be reported honestly as a dataset effect.

Primary Evaluation Order
------------------------

1. Dataset identity and deterministic execution.
2. Blocker-prone seat-share difference.
3. Row-zone concentration control.
4. Multiple-blocker events and temporary aisle reoccupations.
5. Confirmed regions and maximum simultaneous regions.
6. Genuine priority competitions.
7. Cluster selections and starts.
8. Complete cabins and seated totals.
9. Residual stall families.
10. Simulation ticks as a secondary measure.

Diagram Plan
------------
A later diagram may compare neutral seat selection with deterministic
window/deep-seat concentration across dispersed rows. It must show that seat-event
rules remain unchanged and must not depict blockers, dependency regions or
priority competitions as inserted directly by the dataset generator.


======================================================================
RUNNING CONCLUSIONS AFTER EXPERIMENT 2
======================================================================

1. The frozen architecture remains the fixed reference implementation.
2. Experiment 1 established the reusable deterministic dataset layer.
3. Experiment 2 generated seven genuine priority competitions through row-zone
   concentration alone.
4. Simultaneous dependencies therefore do not require human behaviour or
   asynchronous execution.
5. Row concentration produced mixed boarding outcomes and fewer confirmed
   regions overall, so it remains a stress component rather than a complete
   benchmark solution.
6. Experiment 3 isolates assigned-seat depth while restoring dispersed row
   selection and standard shuffled queues.
7. Seat-event rules remain controlled; only the deterministic seat-assignment
   distribution changes.
8. No passenger may pass the assigned row or move backwards.

======================================================================
END OF DETERMINISTIC STRESS DATASET RESEARCH LOG v3 - EXPERIMENT 3 DESIGN
======================================================================
