Phase 2a (lever): expedition retreat semantics split

Phase 2 diagnostics named the InterruptElite bracket as the likely
first lever; tier-walking the lethality probe at the matrix cadence
told a different story.  Phase 1's uniform-0% baseline isn't an
elite-bracket calibration issue at all — every tier reads 0% because
the engine's TimedOut contract was being ignored by every expedition
caller.

combat_engine.go:451 says: "Timeout = retreat, not lethal blow.
Caller treats a timeout loss as 'fight ended, no character death'".
But runHarvestInterrupt / tryPatrolEncounter / resolveCombatRoom all
called abandonZoneRun + retireAllRegionRuns on any !PlayerWon —
ending the expedition outright on a retreat.  The retreat flavor line
("X outlasts you. You retreat from the expedition, wounded but alive")
was already in the code, just stapled to an actual run-abort.

Splits the policy by caller:

  • runHarvestInterrupt   — autopilot daytime interrupt.  TimedOut →
    retreat: threat +5, HP carries over, run continues, harvest slot
    forfeit (no kill / loot).  HP<=0 still ends the run + marks dead.
  • tryPatrolEncounter   — !advance pre-room patrol roll.  Same
    retreat policy: patrols don't gate progress, so retreating from
    one and walking into the next room is the right shape.
  • resolveCombatRoom    — !advance room/elite combat.  Unchanged —
    this path gates room progression; a retreat has nowhere to go, so
    any loss still ends the run.  (Manual zone runs were always
    intended to end here.)

Harness mirrored: daytime interrupt timeout → carry HP + threat bump
+ continue day; night-encounter loss → terminate (mirrors
resolveCombatRoom, since live night encounters defer to !advance).

retreatThreatBump = 5 is the per-retreat threat penalty.  Low enough
not to compound brutally with chained retreats, high enough that 3–4
retreats noticeably walks the threat clock toward Stirring.  Easy to
dial in Phase 3 if zones go off-band.

Phase 1 matrix after the change still reads 0% completion at every
cell — but the encounter counts and survival shape are dramatically
different (T4 underdark 3.6→7.5 encs; T3 underforge trial saw 18
encounters across 10 days where the pre-change run died on day 2/3).
Adds TestExpeditionBalance_Phase2_TierLethality, a tier-walking
companion to the T1/rolls=1 probe, that traces every fight at the
matrix cadence across one zone per tier — the actual Phase 2b lever
work picks from this data, not the old T1-only probe.

The remaining 0% is now legibly driven by tier-disproportionate elite
rosters (Hobgoblin Warchief at T1, Green Hag at T2, Roper/Helmed
Horror higher up) that one-shot or two-shot tier-appropriate
fighters.  Phase 2b's lever shortlist:

  1. Roster gate / SpawnWeight tuning to dilute over-tier elites.
  2. Surprise-nick floor reduction on chained interrupts (carryover
     HP + nick is the death-spiral fingerprint at T1 specifically).
  3. Per-day cadence reduction if 1+2 don't carry T1 to band.

Pre-existing test failures (TestAdv2Scenario_ZoneRunGoblinWarrens,
TestMageSpellbookLineInRender) verified to fail identically on HEAD;
no new test regressions.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

internal/plugin/dnd_expedition_combat.go

@@ -30,6 +30,23 @@ import (
 	"maunium.net/go/mautrix/id"
 )
 
+// retreatThreatBump is the threat penalty applied when the player times
+// out of a combat (PlayerEndHP > 0 but PlayerWon == false). Retreats
+// represent the player breaking off wounded — the run continues, but the
+// zone's awareness ratchets up. Tuned alongside the expedition-difficulty
+// pass (see gogobee_expedition_difficulty.md): low enough not to compound
+// brutally with chained retreats, high enough that 3–4 retreats walks the
+// threat clock toward Stirring.
+//
+// History: pre-Phase-2 the engine's timeout-loss path called
+// abandonZoneRun + retireAllRegionRuns, ending the expedition outright
+// despite the engine's contract ("Timeout = retreat, not lethal blow").
+// That made any single fight loss across a 14-day expedition an
+// auto-fail, which the sim harness exposed as uniform-0% completion
+// across every tier. Splitting the retreat path here was Phase 2's
+// first lever.
+const retreatThreatBump = 5
+
 // ── Combat Interrupt (§4.2) ─────────────────────────────────────────────────
 
 // CombatInterruptKind is the bucket the d20+tier roll lands in.
@@ -146,21 +163,27 @@ func (p *AdventurePlugin) runHarvestInterrupt(
 	}
 
 	if !result.PlayerWon {
+		if result.TimedOut {
+			// Retreat: fighter broke off wounded but alive. The engine's
+			// contract (combat_engine.go: "Timeout = retreat, not lethal
+			// blow") means HP stays where the engine left it and we keep
+			// the run going. The harvest slot is forfeit (no kill, no
+			// loot) and threat ticks up.
+			_ = applyThreatDelta(exp.ID, retreatThreatBump, "combat retreat")
+			b.WriteString(fmt.Sprintf("⏳ **%s** outlasts you. You break off, wounded but alive. (Threat +%d.)",
+				monster.Name, retreatThreatBump))
+			return b.String(), false
+		}
+		// True death.
 		_, _ = applyMoodEvent(run.RunID, MoodEventPlayerDeath)
 		_ = abandonZoneRun(userID)
 		_ = retireAllRegionRuns(exp)
-		if !result.TimedOut {
 		markAdventureDead(userID, "expedition", zone.Display)
-		}
 		if line := flavor.Pick(flavor.PlayerDeath); line != "" {
 			b.WriteString(line)
 			b.WriteString("\n")
 		}
-		if result.TimedOut {
-			b.WriteString(fmt.Sprintf("⏳ **%s** outlasts you. You retreat from the expedition, wounded but alive.", monster.Name))
-		} else {
 		b.WriteString(fmt.Sprintf("💀 You fell to **%s**. Run ended.", monster.Name))
-		}
 		return b.String(), true
 	}
 
@@ -390,21 +413,30 @@ func (p *AdventurePlugin) tryPatrolEncounter(
 
 	var ob strings.Builder
 	if !result.PlayerWon {
+		if result.TimedOut {
+			// Retreat — see retreatThreatBump doc. Run continues; threat
+			// ticks; the patrol's awareness lingers as a soft penalty
+			// instead of an auto-fail.
+			_ = applyThreatDelta(exp.ID, retreatThreatBump, "patrol retreat")
+			ob.WriteString(fmt.Sprintf("⏳ The patrol drags on. You break off, wounded but alive. (Threat +%d.)",
+				retreatThreatBump))
+			if rollLine := dndRollSummaryLine(result); rollLine != "" {
+				ob.WriteString("\n")
+				ob.WriteString(rollLine)
+			}
+			outcome = ob.String()
+			ended = false
+			return
+		}
 		_, _ = applyMoodEvent(run.RunID, MoodEventPlayerDeath)
 		_ = abandonZoneRun(userID)
 		_ = retireAllRegionRuns(exp)
-		if !result.TimedOut {
 		markAdventureDead(userID, "patrol", zone.Display)
-		}
 		if line := flavor.Pick(flavor.PlayerDeath); line != "" {
 			ob.WriteString(line)
 			ob.WriteString("\n")
 		}
-		if result.TimedOut {
-			ob.WriteString("⏳ The patrol drags on. You break off and retreat, wounded but alive.")
-		} else {
 		ob.WriteString("💀 The patrol takes you down. Run ended.")
-		}
 		if rollLine := dndRollSummaryLine(result); rollLine != "" {
 			ob.WriteString("\n")
 			ob.WriteString(rollLine)

internal/plugin/dnd_zone_cmd.go

@@ -755,6 +755,12 @@ func (p *AdventurePlugin) resolveCombatRoom(userID id.UserID, run *DungeonRun, z
 		}
 	}
 	if !result.PlayerWon {
+		// resolveCombatRoom gates room progression — a retreat here has
+		// nowhere to go (the elite/room is still blocking the way), so
+		// any loss ends the run. The retreat-continues semantic only
+		// fits the non-gating expedition paths (runHarvestInterrupt,
+		// tryPatrolEncounter); see retreatThreatBump in
+		// dnd_expedition_combat.go.
 		_, _ = applyMoodEvent(run.RunID, MoodEventPlayerDeath)
 		_ = abandonZoneRun(userID)
 		// Timeout loss = retreat; player took wounds but isn't actually

internal/plugin/expedition_balance.go

@@ -269,6 +269,22 @@ func (h *expeditionHarness) advanceExpeditionOneDay() expeditionTrialResult {
 			res := h.runHarnessFight(zone, kind == InterruptElite)
 			h.encounters++
 			if !res.PlayerWon {
+				if res.TimedOut {
+					// Retreat — mirrors the live retreat semantics in
+					// dnd_expedition_combat.go (Phase 2). Run continues
+					// with carryover HP and a threat bump; the harvest
+					// slot's loot is just forfeit.
+					h.exp.ThreatLevel += retreatThreatBump
+					if h.exp.ThreatLevel > 100 {
+						h.exp.ThreatLevel = 100
+					}
+					h.char.HPCurrent = res.PlayerEndHP
+					if h.traceFight != nil {
+						h.traceFight(fmt.Sprintf("retreat day=%d hp=%d threat=%d",
+							h.exp.CurrentDay, h.char.HPCurrent, h.exp.ThreatLevel))
+					}
+					continue
+				}
 				return h.terminate("died_combat", false, true, false)
 			}
 			h.char.HPCurrent = res.PlayerEndHP
@@ -286,8 +302,10 @@ func (h *expeditionHarness) advanceExpeditionOneDay() expeditionTrialResult {
 	}
 	switch nc.Outcome {
 	case NightOutcomeStandard, NightOutcomeElite, NightOutcomeAmbush:
-		// Live path defers these to !advance; the harness resolves
-		// them inline so the night threat actually pressures the run.
+		// Live path defers these to !advance, where they resolve through
+		// resolveCombatRoom — which ends the run on any loss because
+		// the next-room path is gated on a win. Harness mirrors: a
+		// failed night encounter terminates the trial.
 		res := h.runHarnessFight(zone,
 			nc.Outcome == NightOutcomeElite || nc.Outcome == NightOutcomeAmbush)
 		h.encounters++

internal/plugin/expedition_balance_test.go

@@ -428,6 +428,73 @@ func TestExpeditionBalance_Phase2_LethalityProbe(t *testing.T) {
 	}
 }
 
+// TestExpeditionBalance_Phase2_TierLethality is the tier-walking
+// companion to the T1-only lethality probe. Phase 1's matrix shows
+// uniform 0% across every tier at the default rolls=4 cadence, which
+// the T1/rolls=1 probe alone can't explain (its bimodal-warchief
+// finding wouldn't show at T5 vs an L17 fighter). This probe traces
+// every fight at the matrix cadence across one zone per tier so we
+// can read whether the deaths are elite-driven, standard-fight
+// chained, or something the harness combat fold itself is doing.
+//
+// Diagnostic-only — no assertions.
+func TestExpeditionBalance_Phase2_TierLethality(t *testing.T) {
+	if testing.Short() {
+		t.Skip("phase 2 tier-lethality probe writes verbose log; -short skips it")
+	}
+
+	const trialsPerTier = 3
+	const baseSeed uint64 = 0x7E11A1
+	cells := []struct {
+		zone ZoneID
+		tier ZoneTier
+	}{
+		{ZoneGoblinWarrens, ZoneTierBeginner},
+		{ZoneForestShadows, ZoneTierApprentice},
+		{ZoneUnderforge, ZoneTierJourneyman},
+		{ZoneUnderdark, ZoneTierVeteran},
+		{ZoneDragonsLair, ZoneTierLegendary},
+	}
+	for _, cell := range cells {
+		level := phase1TierCenterline[cell.tier]
+		profile := expeditionBalanceProfile{
+			ZoneID:             cell.zone,
+			Class:              ClassFighter,
+			Level:              level,
+			Supplies:           makeSupplies(cell.tier, SupplyPurchase{StandardPacks: 3}),
+			CampType:           CampTypeStandard,
+			HarvestRollsPerDay: 4, // matrix default, not the sparse probe
+		}
+		t.Logf("═══ %s T%d L%d Fighter rolls=4 ═══", cell.zone, cell.tier, level)
+		for trial := 0; trial < trialsPerTier; trial++ {
+			exp := newHarnessExpedition(profile)
+			char := buildHarnessCharacter(classBalanceProfile{
+				Class: profile.Class,
+				Level: profile.Level,
+			})
+			t.Logf("─── %s trial %d: hp_max=%d ───", cell.zone, trial, char.HPMax)
+			h := &expeditionHarness{
+				exp:         exp,
+				char:        char,
+				rng:         newHarnessRNG(baseSeed + uint64(trial) + uint64(cell.tier)*101),
+				rollsPerDay: profile.HarvestRollsPerDay,
+				traceFight: func(line string) {
+					t.Logf("  %s", line)
+				},
+			}
+			for {
+				res := h.advanceExpeditionOneDay()
+				if res.EndedReason != "" {
+					t.Logf("END %s trial %d: reason=%s days=%d threat=%d encs=%d hp_left_pct=%.1f%%",
+						cell.zone, trial, res.EndedReason, res.DaysElapsed, res.ThreatAtEnd,
+						res.CombatEncounters, res.HPRemainingPct*100)
+					break
+				}
+			}
+		}
+	}
+}
+
 // joinZones is a tiny helper kept local to the test file so the
 // per-tier log line reads in one logical chunk without pulling in
 // strings.Join's import for production code.