Status/turn consolidation pass: tick-scaled decay, generalized trample, action order #201

Follow-up from nnunley's review on #166, tracking the parts of that feedback that are real but out of scope for a narrow bugfix.

1. Status ttl decrements per action, not per tick

add-ticks scales :ticks/regen-ticks by cost but unconditionally calls tick-entity-statuses once per action regardless of that cost. So ttl is documented as "turns remaining" but actually means "actions remaining" — swimming ((* move-cost 2)) burns one ttl same as a normal step, and a speed-modified creature decays at a rate decoupled from the ticks it actually spends.

This is what stands between us and meaningful haste: a player turn would need to cost some multiple of ticks (e.g. ~1000) for a speed effect to matter, and that only works if decay is counted in projected ticks rather than actions.

There's also a structural asymmetry contributing to bugs like #161: the player ticks statuses inside add-ticks, but creatures tick at the top of creature-turn before acting — one concept, two homes.

Note :webbed specifically should not become a projected-time timer — its ttl is a struggle counter (per #71), not a duration, and converting it would let a web time out just by standing still. Any consolidation pass needs to model both counters and timers, or :webbed becomes a timer on the way through.

2. Generalize trampling beyond webs

Trample logic currently lives only in the web struggle branches (world/player-move, ai/try-move). Should have a single handler for "an entity's action affects a hazard tile" covering grass, water, etc., not just webs, and should guarantee at most one entity affects the object per action.

3. Action order / determinism (already filed, folding in here for visibility)

#187: process-creatures (ai.lg:424) takes creature acting order straight from (vals (:entities world)) with no sort, and since every creature action advances the det seed chain, acting order is part of the determinism contract. find-target already guards the same dependency by sorting on (juxt dist-to (comp str :id)); process-creatures doesn't.

Context

Investigation (in #166) confirmed :webbed was the only status with two ttl owners — the other 11 apply/remove sites all go through the single tick-entity-statuses path — so the double-decrement bug itself didn't need a systemic fix. But the tick-scaling and trample-generalization issues above are real and worth doing before anything (like haste) depends on tick-accurate decay.

🤖 Generated with Claude Code