How Tracer itself delivers

collected 1 hour ago, 2026-10-27T14:35Z; provisional

Trailing 7 days to 2026-10-27T14:35Z, withtracer.com from vorthruna/tracer-next. Promotion workflow measures; production incidents and customer outcomes are not measured here. Earlier rerun attempts are assumed failed; their outcomes were not collected. Every figure here is provisional: the measures and their instruments are still being settled, so read each one with the counts and the instrument beside it.

MeasureReadingCountsInstrument
Successful promotions13 (1.9 a day)13 successful promotionspromote.yml success
Merge to promotionmedian 1.9 h, p90 9.5 hn=30 of 30 feat/fix mergesmerge to first promotion after its release
Estimated failed promotion attempts1 of 14 promotion attempts (7%)1 failed attempts, 14 attemptsestimated failed attempts over completed-run attempts; earlier rerun attempts assumed failed, outcomes not collected
Run start to successmedian 0.0 hn=1, 0 without a later successa failure to the next success, or a rerun from its first attempt to its final success
Main build failures17 of 91 push runs17 red, 91 runsci.yml on main, event push

By surface

A merge counts in every surface its changed paths touch: data is migrations/, web the public pages, app the rest of src/.

SurfaceMergesPromotionsMerge to promotion mediann
app1381.4 h13
web431.5 h4
data00not enough evidence: no feat/fix merge touching data in windownone

Changes with no test touched

A feat/fix merge counts when it reached production and none of its changed paths is under test/. A merge that changed documentation alone (docs/, Markdown files) is set aside: it has nothing a test could prove. It says nothing about whether the tests that exist were enough.

Each link opens the commit on GitHub, for collaborators of the repository; anyone else gets a not-found page.

Sources

Programs read the same document at /api/delivery/dora.