Release verdict maps in-flight builds to "failed", making a healthy factory read as all-red

[castrojo]

Summary

data/release-verdict.json reports build.status: "failed" for builds that are still running, with the reason string stating so outright. The dashboard then propagates that into the verdict cards and into "What's degrading?", so https://factory.projectbluefin.io/ currently reads as a total outage while the factory is in fact healthy.

Evidence

From https://factory.projectbluefin.io/data/release-verdict.json, live:

lane	verdict	build.status	finished_at	build.reason
bluefin-stable	bad	failed	null	latest publishing run concluded running
bluefin-testing	bad	failed	null	latest publishing run concluded running
bluefin-lts-stable	bad	passed	set
bluefin-lts-testing	pending	passed	set
dakota-testing	bad	failed	set	latest publishing run concluded pending

Three of five lanes are marked failed for not having finished yet. Two of them have finished_at: null, which is a direct signal that the run is still in flight.

Meanwhile the same page's own rolling-7-day panel reports 79 builds, 74 passed, 5 failed, 94% success rate over the same window.

Impact

• 0 of 5 release lanes render "good", so the page reads as a disaster area.
• The error compounds: "What's degrading?" restates each false failure, so one mapping defect produces roughly six alarming rows.
• This is the single largest reason the dashboard looks broken to a visitor.

Suggested fix

Map GitHub Actions conclusions explicitly, and treat a non-terminal run as pending rather than failed:

• conclusion === "success" → passed
• conclusion === "failure" | "timed_out" | "startup_failure" → failed
• status !== "completed", or conclusion null/running/pending, or finished_at === null → pending

ADR 0002's principle that a lane is never inferred good is right and should stand; the inverse also needs to hold, so a lane is never inferred bad purely from incompleteness.

Worth adding a regression test asserting that a run with finished_at: null never yields build.status: "failed".

Context

Found while auditing both dashboards ahead of consolidating project-level status onto https://docs.projectbluefin.io/factory. Filing here since the mapping lives in this repo.