docs(dev): remove cap counts and line numbers from the in-flight census
What this change does
Two documents describe one family of hand-rolled in-flight caps.
Both state how many caps exist.
docs/dev/storage-accounting.md also tables each cap with a file:line coordinate.
The count and the coordinates both rot, and both had rotted already.
At 7307bd94a, three of the eleven tabled coordinates did not match the census command.
This change removes every stated count of caps from docs/dev/storage-accounting.md and docs/dev/remote-read-budgets.md.
It removes the :NNN from every row of the three census tables and keeps the constant and the path.
The census command git grep -n "MaxInFlight = 64" -- internal stays.
So does the sentence that says the line number is whatever the command reports.
Each partition heading opened with its count, so three anchors change name and three inbound links move with them.
One sentence in internal/format/oci/emit_dispatch.md said the census "tables every one by constant and file:line".
The coordinate removal makes that sentence false, so this change corrects it to say constant and path.
A second sentence in the same file, at :38, told a reader to identify a cap by its path alone.
The coordinate removal makes that key ambiguous, because one path is the "Declared at" value of two rows in two partitions, so this change names both halves of the key there and at docs/dev/storage-accounting.md:1490.
The diff is 52 insertions and 51 deletions across three files.
Every hunk replaces one line with one line, except one sentence added to docs/dev/remote-read-budgets.md to say which set its exhaustiveness claim is over.
Documents this change does not touch
Three documents under docs/plans/ carry the same falsified prose and stay out of this diff on purpose:
docs/plans/2026-09-03-repository-column-writers.mddocs/plans/2026-09-03-upload-session-concurrency-cap.mddocs/plans/README.md
Work item #1197 records all three, and a separate docs(plans): merge request repairs them.
docs/runbooks/npm-packument-cache-invalidation.md also tables a coordinate, and the issue excludes that file by name.
Its coordinate is correct at 7307bd94a.
Merge order
!2324 and !2362 have both merged, and this branch is rebased onto both census rewrites.
99ebd6be2 added a twelfth cap.
c8784251d added a thirteenth, repositoryStampMaxInFlight in internal/format/oci/emit_dispatch.go.
Each one raised every stated count and renamed a partition heading, and this change removes what they added.
Measured at head 24a469530, whose base is b5287abe3: the census tables carry 13 rows.
The base carries the same 13, so no count here is owed to a later rewrite.
All six inbound links resolve to the three renamed headings.
The earlier reading of eleven rows belonged to the pre-rebase base 7307bd94a.
main advances on its own, so a behind count is a property of the reading rather than of this change.
!2262 is open at 6f267b27c and still shares two of this change's three files.
git merge-tree between the two branches puts docs/dev/storage-accounting.md and internal/format/oci/emit_dispatch.md in the conflict set.
The first of the two to merge puts both files into conflict for the other.
Whoever resolves that conflict takes !2262's link retarget together with this change's coordinate removals.
That resolution re-derives every line number at its own head, because this description states none.
Verification
No Go test applies, because this change edits Markdown only.
The gate is lychee --offline --include-fragments over **/*.md, which blocks in the docs:lint CI job and in a pre-commit hook.
At the branch head it checks 8438 links repo-wide and returns 0 errors.
A negative control ran as well.
One moved link went back to its old anchor in a scratch copy, and the same command returned exactly one Cannot find fragment error.
The gate therefore detects this class of defect, and a clean run is evidence rather than silence.
markdownlint-cli2 returns 0 issues over 235 files, and vale --minAlertLevel error returns 0 errors over 410 files.
End-to-end scenarios
This change affects no scenario.
It edits documentation, it changes no behavior, and docs/testing/ needs no new entry.
One judgment call in the diff
The issue protects one sentence by name: "A cap is identified by its constant and its path, and the line number is whatever the command reports." That clause survives word for word. The clause after it read "so a coordinate that has drifted is read off a fresh run rather than off a stamp". No coordinate remains in that section, so "a stamp" lost its referent. The clause now reads "so the tables below stamp no coordinate and a reader who wants one runs the command". This is a judgment about the issue's wording rather than about the tree. This description names it, so a reviewer does not have to find it.
The issue's own instructions carry sixteen items, and this branch has evidence for all sixteen.
Five further changes are derived rather than stated by the issue, so this merge request carries Related to.
The issue states the count removal; the wording that replaces a removed count is not stated, and these five are that wording.
They are the second claim of the Known-gaps row at docs/dev/storage-accounting.md:1465, the cap-repetition clause of that same row, the rewrite of docs/dev/storage-accounting.md:1488, the rewrite of internal/format/oci/emit_dispatch.md:38, which the issue scopes to one line at :36, and the exhaustiveness scoping at docs/dev/remote-read-budgets.md:176 with the two lines that carry it at :178 and :179.
All five are met, so the form names who set the requirement rather than any remaining work.
Related to #1193
This is a bot message