[mikaeltarquin]

Summary

The automatic check weighed all of a saved item's picks together. A freeleech audiobook and a small ebook waited as long as a large non-freeleech audiobook picked with them couldn't go, which wasted free unsatisfied slots that would have been building ratio.

• Each pick is now weighed on its own, cheapest first: nothing on MAM, then freeleech (still checked live), then by size. Each one goes if it fits (unsatisfied slots, buffer, target ratio) alongside the picks already going.
• Picks that don't fit stay on the item in the queue with "Waiting for" saying why. They stay queued, unlike a pick that failed to queue, which still turns the item off with last_error.
• SavedItemsService.keep_picks gains a "still waiting" mode for this: keeps auto_get, sets auto_status, no error.
• The Queued page's description now says each release is decided on its own.

Testing

• New tests:
  • Freeleech audiobook and small ebook go while a 40 GB non-freeleech audiobook waits for the ratio.
  • With one free slot, the freeleech pick gets it ahead of a paid pick listed first.
  • A partly downloaded item stays queued with its remaining picks.
• The new and updated tests fail on the old code.
• Python (-m "not integration and not e2e"): only the same 5 environment failures as before.
• Frontend: typecheck, oxlint, oxfmt, vitest (348).

🤖 Generated with Claude Code

https://claude.ai/code/session_01V88DbAixT45r3AJsyFJchY

---

Generated by Claude Code