docs: file #429 — periodic local-player run hitch; GC/pack exonerated, prediction-vs-ACE speed snap is the lead

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/ISSUES.md

@@ -24,6 +24,56 @@ What does NOT go here:
 - Every session: scan OPEN issues at start; promote/close anything we touched during the session before ending.
 - Promoting to a Phase: mark as `DONE (promoted to Phase X)` + commit SHA where the Phase entry landed.
 
+## #429 — Periodic hitch in the LOCAL player's own motion while running (~2-3 s cadence)
+
+**Status:** OPEN
+**Severity:** MEDIUM (noticeable during ordinary play)
+**Filed:** 2026-08-23 (owner report during the night-sky session)
+**Component:** movement / prediction / server reconciliation
+
+**Symptom:** every ~2-3 seconds while RUNNING forward, the player's own
+motion visibly hitches ("small lag, like spike"). Not present when walking
+or strafing. Birds and every other world object stay perfectly smooth
+through the hitch — so it is NOT a frame-time spike; it is the local
+player's position stepping.
+
+**Evidence gathered at filing:**
+- 45 s `dotnet-counters` capture while reproducing (Atmospheric pack ON):
+  Gen0 GC every ~6.3 s (cadence mismatch), gen1/gen2 zero, %time-in-GC ~0
+  — GC exonerated for the 2-3 s cadence. (Separate finding: the pack path
+  allocates a steady ~8 MB/s ≈ 130 KB/frame where the base client runs
+  near zero — worth its own cleanup issue.)
+- Pack GPU-timer readback is non-blocking by design (VulkanGpuTimerPool
+  tolerates NotReady) — not a stall source.
+- The owner initially observed "pack off → gone"; after the walk/strafe/
+  world-smooth discriminators this is more plausibly a VISIBILITY effect
+  (lower FPS with the pack makes a 2 s position snap read as a lag spike).
+
+**Leading hypothesis:** client-predicted run speed vs ACE-authoritative
+speed mismatch at high runRate (the test character applies server stats
+run=15230 — deep in the ==800-sentinel family, see #266: retail's ==800 vs
+ACE's >=800 misread), accumulating drift that the AutoPos/correction loop
+snaps back on its cadence. Walking (speed 1.0 both sides) and strafing
+accumulate little or no drift — matching the report exactly.
+
+**Next probes (in order):**
+1. `ACDREAM_DUMP_MOTION=1` + a temporary inbound-position log for the
+   LOCAL guid: does ACE send position sets for the local player every
+   ~2-3 s while running, and does acdream apply them? (Retail ignores
+   routine UpdatePosition for the autonomous local player; only
+   ForcePosition snaps — verify our projection split here.)
+2. Compare `PlayerWeenie.InqRunRate(15230)` against ACE's computed
+   runRate for the same skill — a formula delta is the root cause if the
+   correction traffic confirms.
+3. If no corrections arrive: instrument the local movement controller's
+   own periodic state (autorun latch, tracker send side-effects).
+
+**START at** `claude-memory/project_physics_collision_digest.md` (per-cell
+DO-NOT-RETRY rules) and `claude-memory/project_retail_motion_outbound.md`
+(TS-33 AutoPos tracker semantics) before instrumenting.
+
+---
+
 ## #428 — Sky lightning slot fired 3 CreateParticle hooks with no live pose after a TimeSync day-group flip
 
 **Status:** OPEN