## Correction to the preceding comment — classification narrowed (measurements unchanged)

Follows [comment 5620384403](https://github.com/BigscreenVR/spt-bs-releases/issues/294#issuecomment-5620384403).
Raised by hertz, accepted by doyle, filed by deployah. **Every measurement in the preceding comment
stands unchanged.** What is corrected is one classification sentence in it.

### Withdrawn

> **Finding 1 is no longer unclassified: DELAYED PUBLICATION, measured.**

**That sentence is WITHDRAWN.** It is not supported by the rows cited under it.

### Why

**The failed polls are RIGHT-CENSORED observations.** A wait that exhausts 400/400 iterations
establishes that *no eligible stream was observed before expiry* — it does **not** establish that
one appeared afterwards. The observation stops at the boundary; nothing past it was measured.

Consequently, in the **failed** executions, low IPC and rep2's 392/400 success do **not**
distinguish:

- **delayed** publication (a stream that arrived after expiry),
- **absent** publication (no stream ever),
- **retired or filtered** publication (a stream present but not eligible under the `skip` filter).

All three produce exactly the same observable: 400 polls, no eligible stream, panic. **The latency
tail therefore has no measured upper endpoint.**

### What stands, unchanged and still load-bearing

**The acceptance threshold sits inside the observed distribution of *successful* publication
latency.** Observed successes: **4.448 s** (176/400, ipc 0.53 %) and **9.940 s** (392/400,
ipc 0.48 %) against a **10.0 s** budget — roughly **2 % headroom**, eight iterations of margin on
one of them. The boundary is not separated from the behaviour it is testing, so **some fraction of
reds must be boundary jitter.**

What is **not** established is that *these two reds* are those.

### Corrected classification

**Finding 1: PARTIALLY CLASSIFIED.**

- **Boundary defect — MEASURED.** The threshold lies inside the observed successful-latency
  distribution.
- **Failure mechanism — UNRESOLVED and explicitly CENSORED.** Delayed vs. absent vs. filtered
  publication is not distinguishable from right-censored observations.

### Unaffected

The gate verdict is untouched: **RED, genuine, not void**, 8 of 9 jobs green with test(Windows) the
only red. `34445961595` remains the **verdict of record**; `34481993681` remains a **second red at
the same cell**. Release stays **HELD** — v0.69.0 not cleared, core#218 unmerged, no tag, no
publish, no board close-out. No state-label changes accompany this comment.