[sdubois777]

The dungeon's floor rules now count the time each beat actually covered, up to one second, instead of taking every beat as 0.25 s. A slow frame no longer slows them, and a long stall costs at most one second of a rule's effect (#1613).

What changed

• game/Source/Cataclysm/Dungeon/CataclysmDungeonGameMode.h and .cpp, the dungeon game mode: Tick records BeatSeconds, the time the beat covered, before the wave-check accumulator resets, and clamps it to 1.0 s. The 79 uses of the fixed 0.25 s in the floor rules read it instead. The waves are unchanged.
• game/Source/Cataclysm/Tests/CataclysmDungeonModifierEffectsTests.cpp: three tests -- Mortal Decay's rate is the same at four and at two frames a second, a three-second frame costs one second, and Death's Embrace's clock counts the same at both frame rates.
• docs/DECISIONS.md: the entry and window w77's record.

Evidence: window w77, at 3f6b0af on development c120310

• Build: 72 actions, 62 files compiled.
• The three new tests alone: 3 performed, 3 succeeded. Printed: "a frame of one second cost 5.100 and a frame of three cost 5.100"; "Mortal Decay over one second cost 5.100 at four frames a second and 5.100 at two"; "Death's Embrace cut 0.0 then 10.0 at four frames a second, 0.0 then 10.0 at two".
• Whole Unreal suite: 4347 performed, 4347 succeeded, 0 failed (registered 4347; 4347 declared in the tree, gap 0). Memory: peak 8,086 MB, final 7,777 MB.
• Python: "7022 passed, 8 skipped"; JUnit tests=7030, failures=0, errors=0, skipped=8. ruff: "All checks passed!".
• Proof P1 (BeatSeconds set back to 0.25 s): PROVED. With the break in, ABeatCountsItsOwnTimeSoMortalDecaySapsItsRateAtTwoFramesASecond failed; restored, it passed.
• Proof P2 (the one-second limit removed): PROVED. With the break in, ABeatCountsAtMostOneSecondSoALongFrameCostsOneSecondOfMortalDecay failed; restored, it passed.

After the tested commit, only docs/DECISIONS.md changed (the record).

🤖 Generated with Claude Code