repro: a panicked scan worker leaves the scanner busy forever

[dorianvp]

Claim

A panic inside scan(..) leaves a ScanWorker busy forever, so the scanner ticks without progress and without an error.

• pepper-sync/src/scan/task.rs:660-690: ScanWorker::run spawns a tokio task and stores the JoinHandle in self.handle, but nothing ever polls it. is_scanning is stored false only at the end of the loop body, after scan(..) returns.
• pepper-sync/src/scan/task.rs:585: Loader::check_error shows the pattern that is missing for workers.
• pepper-sync/src/scan/task.rs:169: Scanner::idle_worker only returns workers with is_scanning == false.
• pepper-sync/src/scan/task.rs:252: ScannerState::Shutdown only drains idle workers.
• pepper-sync/src/scan.rs:142-154: scan panics on .expect("compacts blocks should not be empty") and on panic!("compact blocks do not match scan range!"), which are reachable from server data.

What the test does

scan::task::tests::panicked_scan_worker_is_reported_or_idle builds a ScanWorker with MainNetwork params, empty ufvks, and an unbounded results channel, calls run(1024), and sends a ScanTask for range 10..20 with no compact blocks. After 500 ms it asserts the invariant that a panicked worker is either reported (an Err on the results channel) or marked idle.

Observed output on current dev

thread '...' panicked at pepper-sync/src/scan.rs:144:10:
compacts blocks should not be empty

thread '...' panicked at pepper-sync/src/scan/task.rs:895:9:
a panicked scan worker must be reported or marked idle, but is_scanning=true and no result was sent

test result: FAILED. 0 passed; 1 failed

The worker task is finished (handle.is_finished() == true), the JoinError is never observed, is_scanning stays true, and nothing is sent on the results channel.

Notes

This PR is a reproduction only and does not contain a fix. The sync-bench A/B rule was not run because the commit adds a test only.

🤖 Generated with Claude Code