Bound watch_for_import so stuck transfers can't stall downloads (fixes #30)

[bugrax]

Fixes #30.

Problem

watch_for_import polls is_imported() in an unbounded loop and only exits once every video file of a transfer is imported:

loop {
    if transfer.is_imported().await { /* clean up, send Imported */ break; }
    sleep(polling_interval).await;
}

If a transfer contains a video file the *arr will never import — most commonly a sample that isn't inside a skip_directories folder — is_imported() is permanently false and the task loops forever. Each such transfer leaves a forever-running task polling the *arr/put.io APIs every polling_interval. They accumulate, and after ~10h with a large library the process stops picking up newly-completed transfers entirely (downloads stall with many COMPLETED transfers left un-pulled on put.io). A restart only clears it temporarily.

The tell-tale sign in the logs is the main file of the same transfer being reported found imported by radarr over and over every ~10s — the movie imported fine, but a sibling sample keeps the transfer from ever completing.

Fix

Give up watching after MAX_IMPORT_WAIT (2h) with a warning, so a stuck transfer ends its task instead of polling forever. Genuine imports are detected within a poll or two of the *arr importing, so this generous bound doesn't affect normal operation; it just caps how long a never-importing transfer can sit in a polling loop.

This is a targeted fix for the stall. The deeper fix is better import-mapping (the existing skip_directories TODO — e.g. ignoring sample files, or treating a transfer as imported once it has left the *arr download queue), which would let these transfers complete and be cleaned up rather than be abandoned.

Note

A given-up transfer's local and put.io copies are no longer auto-cleaned (it stops being tracked). That's a deliberate trade-off — leaking one stuck transfer is far better than stalling every download — and it is re-evaluated on the next restart.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds a bounded wait to prevent watch_for_import from polling indefinitely when is_imported() can never become true (e.g., transfers containing files the *arr won’t import).

Changes:

• Introduce MAX_IMPORT_WAIT (2h) as an upper bound for import polling.
• Track start time with Instant and stop watching after the timeout with a warning.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

[bugrax]

These were addressed in follow-up commits already on the branch:

• removed log now only fires on the actual cleanup/import path, not on timeout (95bc322); the timeout path has its own explicit warn ("giving up watching …"), which is the terminal log for that case.
• The 2h bound is now configurable via import_timeout_secs (9645afb), defaulting at the type level with #[serde(default = "default_import_timeout_secs")] returning 7200 (94206d0), and 0 disables it.

On giving up leaving local/remote copies behind: that's an intentional trade-off — leaking one stuck transfer is far better than the unbounded loop stalling every download — and it's re-evaluated on the next restart. The accompanying log makes it greppable. Open to surfacing it as an explicit failure event if you'd like.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.