feat(35-10): OTEL watcher events for consequence, combatant, party_reconciliation, progression

[slabgorb]

Summary

Instruments four sidequest-game subsystems with OTEL watcher events so the GM panel can verify they are engaged and making decisions:

• consequence — consequence.wish_evaluated event fires when evaluate() is called (both true and false branches per no-silent-fallbacks)
• combatant — combatant.bloodied event emitted from broadcast_state_changes(), gated on friendly characters below 50% HP
• party_reconciliation — party_reconciliation.reconciled event fires when reconcile() completes a merge
• progression — progression.stat_scaled event for all three scaling functions (level_to_hp, level_to_damage, level_to_defense)

Implementation Details

Rework History

Initial approach (reviews #1–#2): Instrumented combatant.bloodied inside Combatant::hp_fraction() default trait method.
Rejection root cause: The method had zero non-test production callers — state.rs::lowest_friendly_hp_ratio() and broadcast_state_changes() both inline the fraction math instead of delegating.

Final approach (rework #3, Architect Option A): Relocated combatant.bloodied emission to state::broadcast_state_changes() where the actual state transition is observed. Reverted hp_fraction() to a pure accessor. This restores the OTEL pattern: "emit at mutation sites, not inside pure accessors."

Test Status

• 19/19 OTEL watcher event tests GREEN
• 487/487 sidequest-game integration tests GREEN
• Behavioral wiring test verifies broadcast_state_changes() → OTEL channel (not source-grep)
• All four subsystems validated to emit events

Design Deviations

TEA (test design)

• Combatant trait instrumented at the `hp_fraction()` default method instead of at production call sites → ✗ FLAGGED by Reviewer: this deviation is the root cause of the rejection. TEA's reasoning was internally consistent, but the assumption that production state-build code uses `hp_fraction()` was never verified. It doesn't. `state.rs::lowest_friendly_hp_ratio()` (line 402) and `broadcast_state_changes()` (lines 797-798, 890-891) all call `Combatant::hp()` and `Combatant::max_hp()` directly and inline the fraction math. The instrumentation TEA placed in `hp_fraction()` is decorative — it's reachable from tests but not from any production code path. The wiring assertion TEA wrote checked for `Combatant::hp(` strings (which exist in state.rs for unrelated reasons), giving false confidence. Rework required: either move the instrumentation to where bloodied is actually computed (`state.rs:402`), or change the production code to delegate to `hp_fraction()`. The latter is a 2-line fix and is the recommended path.

  • Severity: minor → HIGH (escalated by Reviewer) — the assumption that callers go through `hp_fraction()` is false in the existing codebase, making the instrumentation unreachable.

• Single `progression.stat_scaled` event for all three scaling functions instead of three separate events → ✓ ACCEPTED by Reviewer: matches the 35-9 precedent of grouping related sub-decisions under one event component with a discriminant field. All three scaling functions are correctly distinguished by the `stat` field.

• `xp_for_level()` is NOT instrumented → ✓ ACCEPTED by Reviewer: correct call. `xp_for_level()` fires on every action via the threshold check — instrumenting it would flood the watcher channel.

• `consequence.wish_evaluated` fires on BOTH power-grab true and power-grab false → ✓ ACCEPTED by Reviewer: textbook application of the no-silent-fallbacks rule.

Dev (implementation)

• combatant.bloodied OTEL emission relocated from `Combatant::hp_fraction()` to `state::broadcast_state_changes()` (rework refactor(game): extract CreatureCore — shared struct for Character and NPC #3) → ✓ ACCEPTED by Reviewer (re-review refactor(game): extract CreatureCore — shared struct for Character and NPC #3): the architect-directed relocation closes the HIGH wiring violation. Production reachability confirmed via grep at `dispatch/mod.rs:1737`. Behavioral wiring test replaces the prior source-grep loophole. Field shape preserved exactly.
  • Severity: major (OTEL emission site moved across files)

Forward Impact

• Story 5-7 wiring debt (`lowest_friendly_hp_ratio` not wired into pacing pipeline) is explicitly NOT closed by 35-10 — separate ticket recommended
• Follow-up: standardize wiring tests across four subsystems (three still use source-grep, only combatant has behavioral test)
• Follow-up: add negative test case for `is_friendly=false` filter in `broadcast_state_changes` (non-blocking improvement per Reviewer re-review refactor(game): extract CreatureCore — shared struct for Character and NPC #3)

---

Fixes MSSCI-35. Approved by Reviewer in re-review #3 with final commit d277562 (all tests GREEN).