Answer A-490k: the indexer's catch-up says take-in.taken with its entries; a runner behind on its view asks again on the next event

Claude-Session: https://claude.ai/code/session_01JG73KHUnBze48rdua8u2sy
This commit is contained in:
2026-11-19 23:03:39 +01:00
parent 0928ed52be
commit ccacffdc90
@@ -2104,3 +2104,10 @@ the principle that settles it, cited as "design 47 G A-n".
build whose recorded digest differs from what the holder holds at its version fails loudly, version taken, and is
never marked released. (Live 2026-11-19 21:09 and 21:19: every publish refused "reserved for another digest", the
reservation's digests empty until the build ends.)
- **A-490k** (A-463, A-464, §493; principles 1 and 3) The indexer's catch-up, which takes a watched branch's heads in
directly when the gate's source names merges it has not taken in, says `take-in.taken` as every other source does:
its branch, source, revision and one entry per module it wrote. A pending merge it covers then ends with no entries,
as today, and starts nothing twice. The pipeline runner whose catalog view is behind a take-in's revision asks the
view again on its next event and runs the starts it held once the view reaches that revision, with no timer. (Live
2026-11-19 22:40: the catch-up took mesh-delivery 0ea4b60d in, writing six modules, and said nothing, so no
pipeline started for them.)