v1.4.0

See CHANGELOG.md for the full release notes.

What's Changed

• The free-threaded canary's Linux hang was dash's empty job table, not the server by @codetalcott in #307
• The tick hook's duty cycle, and a reference for periodic work off the loop (N15) by @codetalcott in #308
• Docs and setup: the counter's bus guard, a SessionStart hook, and a GPU correction by @codetalcott in #309
• A pool-less loop releases the GIL while it waits, so app threads run (#310) by @codetalcott in #311
• A mount set that leaves a mount unserved is refused, not half-started by @codetalcott in #312
• A mount set that cannot run inline keeps the default pool under --workers by @codetalcott in #313
• A path no mount claims gets the server's 404, whatever the first mount is by @codetalcott in #314
• Mojo pool threads register on their lane, so the Mojo mount scales by @codetalcott in #315
• Mounts under --threads get a gate that can pass, run where it can run by @codetalcott in #316
• M0_POOL_SPIN_US is read, as four records have said since 0.19.0 by @codetalcott in #317
• WSGI pool threads register in start, so a stop racing the start ends by @codetalcott in #318
• bench-mojo-mount runs on a fresh clone, and its artifact says what it measured by @codetalcott in #319
• A Mojo mount's compute route runs on all its threads again (SPEC M25) by @codetalcott in #320
• smoke-sim-loop's on-loop arm stops being a lottery on a slow runner by @codetalcott in #323
• m0pub numbers only from a page it has verified, and hands a child the page by fd (#322) by @codetalcott in #324
• Release 1.4.0: changelog, version bump, pre-release gates and re-recorded benchmarks by @codetalcott in #321

Full Changelog: v1.3.0...v1.4.0