# PENGO-E015 — Fixed timing extension v2

The first native batch is terminal:33 recipes,66 fresh processes,330,706 frames,735.047seconds. Its registered key combined input and observed prefixes, yielding20 contact episodes and16 usable timelines. A stricter audit merges actors with identical observed prefixes even when their input histories differ. This yields17 observed contact episodes and14 usable timelines; the three merged groups are retained, not removed from the original receipt. Shared boards and paired replays remain dependent observations.

Complete timelines already exhibit97–121-frame delays. The117-frame seed value is therefore not constant in these tested contexts. Do not replace it with another fixed timer. Counter acknowledgement remains event-driven.

Before outcomes, freeze an extension from all eight original PhaseC roots in original case order. For each root, insert exactly3,4,5 or6 neutral frames before the same two-frame legal directional push, then600 neutral frames.32 recipes, two native reset replicas each, eight processes maximum concurrently, batch wall1800seconds, per-process timeout240seconds and maximum800,000 frames. No root is chosen by score or successful crush. Freeze actual manifest budget before launch. Start only after E016 native egg batch is terminal. Deadline7October07:00UTC; infrastructure/prefix/parity failure stops further jobs with no automatic retry.

Use the existing five event clocks, phase censoring and conservative attribution. Combine with the preserved first native batch, deduplicating by observed contact-prefix hash, ACT and actor rather than input history. Keep every attempt, miss, censored or confounded timeline. The first30 distinct matching contacts in fixed old-then-new recipe order form the declared report sample; retain all later contacts in the complete result. The latency gate requires at least20 complete attributable events in that sample, with two native replicas each. If insufficient, report the gap; do not call recipes, duplicate tapes, replicas or contact-only cases complete events.

Observed-prefix equivalence remains narrower than complete-machine identity. Native trace/pixel/device parity and variable latencies do not establish independent starts, later ACT rules or a full-game record. No objective planner or scoring search executes in this diagnostic.