Capture SHA differs from Session 4 and should be tracked as a new session family.
Capture stream is valid JSONL with 11190 tick rows and no terminal capture marker (expected for current writer behavior).
In this capture family, input_approx.move_dx/move_dy appears to reflect runtime movement deltas, not stable raw input intent; replaying them directly caused a false early drift at tick 61.
Reconstructing replay input from input_player_keys (with fallback merge) moved the frontier from tick 61 to tick 3254.
Current lead stack on this SHA is:
unresolved RNG accounting/branch parity around tick 3050 (perk_select_random caller),
then movement/position drift at tick 3254,
with an opposite-direction key conflict observed in capture key-state at the focus tick.
uv run pytest tests/test_original_capture_conversion.py
uv run crimson original verify-capture artifacts/frida/share/gameplay_diff_capture.json --float-abs-tol 1e-3 --max-field-diffs 32(expected non-zero exit while diverged)
uv run crimson original divergence-report artifacts/frida/share/gameplay_diff_capture.json --float-abs-tol 1e-3 --window 24 --lead-lookback 1024 --run-summary-short --run-summary-short-max-rows 30 --json-out analysis/frida/reports/capture_508bcc_divergence_after_inputkeys_fix.json(expected non-zero exit while diverged)
uv run crimson original focus-trace artifacts/frida/share/gameplay_diff_capture.json --tick 3050 --near-miss-threshold 0.35 --json-out analysis/frida/reports/capture_508bcc_focus_3050_after_inputkeys_fix.json
uv run crimson original focus-trace artifacts/frida/share/gameplay_diff_capture.json --tick 3254 --near-miss-threshold 0.35 --json-out analysis/frida/reports/capture_508bcc_focus_3254_after_inputkeys_fix.json
Use the updated Frida script in the next capture so input_approx carries mirrored key-state consistently.
After that re-capture confirms stable key-state telemetry, remove temporary replay fallbacks in src/crimson/original/capture.py that synthesize/mix input from partial fields.
Then continue RNG/branch investigation at tick 3050 (perk_select_random) before re-evaluating the tick 3254 movement drift.