image reader waits out a registry 429 and tries again (3 tries); 08 + changelog: the dry-run measurement of the Docker Hub change (0 refused), two runners = two tag caches (R-1334)

documentation/architecture/08-alarm-ladder.md

@@ -706,5 +706,8 @@ morning sent 224 and was refused none. So: the requests to Docker Hub are spaced
 a 429 waits (its `Retry-After`, else 10/30/60 s) and tries again, and the image is NOT CHECKED only after the fourth
 try; a tag list is kept for 24 hours per registry path in the CI runner's cache (`~/.cache/felhom-watch-tags.json`, not
 committed), so two names of one path (`lscr.io/linuxserver/x`, `linuxserver/x`) and a same-day re-run fetch nothing.
-The log's `TAGS fetched=… from-cache=… retried-429=…` line counts it. No Docker Hub login is used.
+The log's `TAGS fetched=… from-cache=… retried-429=…` line counts it. No Docker Hub login is used. Two CI runners take turns, each with its own
+cache, so the cache helps a same-day re-run on one runner, not the nightly; the spacing is what fixes the night (two
+dry runs 2026-10-28: 0 refused). The image reader (`image_runtime.py`) also waits out a 429 and tries again (3 tries):
+ghcr.io refused two blob reads when two dry runs read the same 14 new images 15 minutes apart.