AKUNUR 3.1.0-alpha.10 VALIDATION REPORT

Scope
-----
Final planned heuristic-AI alpha for three-player Tactical. No game-rule changes
from 3.1.0-alpha.9. Alpha.7 Still Point route planning, alpha.8 corner strategy
and alpha.9 multiplayer leader awareness are retained.

Focused alpha.10 validation
---------------------------
PASS deadlock-breaker-test
  - fresh productive plans do not trigger deadlock breaking
  - a deeply stagnant plan selects a different safe move relation
  - the stale corner target is reset after intervention
  - deadlock cooldown/private counters are updated
  - urgent leader pressure suppresses the breaker
  - aggregate Deadlock breaks metrics synchronise correctly

Retained multiplayer validation
-------------------------------
PASS multiplayer-awareness-test
  - shared leader recognition
  - leader support disruption
  - own winning-plan guardrail
  - emergency third-corner response
  - leader-pressure retarget
  - integrated support-breaking Leader response
PASS corner-strategy-test
  - ready third-corner conversion
  - sticky corner commitment / sustained-failure retarget
  - opponent two-corner threat targeting
PASS still-point-route-planner-test
  - qualification route commitment
  - supported blocker clearance
  - productive reversible sacrifice
  - qualification passage execution
  - sticky route commitment / passage override
  - objective stagnation counter
PASS three-player-ai-test
PASS multiplayer-progress-test
PASS three-player-manual-test
PASS player-foundation-test
PASS reserve-lifecycle-test

Inherited regression
--------------------
PASS smoke
PASS parity
PASS alpha.11 rules regression
PASS result regression
PASS analysis regression
PASS two-player Tactical regression
PASS two-player Strategic regression
PASS symmetry regression
PASS seeded determinism regression
PASS replay/resume regression
PASS browser module smoke

The two-player Tactical and Strategic agents remain green after the three-player
alpha.10 work.

Performance / search boundary
-----------------------------
- alpha.7 qualification Still Point exit pruning retained;
- alpha.8 corner-awarded Echo placement pruning retained;
- repeated corner-formation analysis remains cached per state revision;
- alpha.9 post-qualification Still Point transit pruning retained;
- alpha.10 deadlock selection reuses the already-scored Tactical candidate set
  and does not add nested opponent or alternative game-tree searches.

A short 80-decision seeded three-player Tactical observation completed legally
with no deadlock intervention before strategic stagnation arose. Longer Lab
analysis runs can be substantially slower than browser autoplay in complex
post-qualification positions, so browser autoplay remains the intended alpha.10
behavioural validation path.

Interpretation
--------------
Alpha.10 is intended to end handcrafted Tactical feature development. Browser
validation should now ask whether Deadlock breaks occur rarely and whether some
of the high-Anti-thrash 400-move alpha.9 games convert into dynamic play without
harming clean wins or alpha.9 leader responses. If so, the next development
step is 3.1 release validation and seeded simulation/strategy discovery.
