The operation did not complete within the given timeout. The spawned
worker thread is left to run to completion in the background (there is
no safe way to cancel a blocked syscall from the outside); this trades a
leaked thread for never hanging the caller.
Default ceiling for any single blocking file/subprocess operation that
touches a path supplied by the caller (e.g. a scanned media file). Chosen
to comfortably exceed a slow spinning disk or network share while still
surfacing a stale/disconnected mount point (which otherwise blocks the
underlying syscall forever on macOS) within one command’s run.
Runs f on a helper thread and waits up to timeout for it to finish.
Use this to bound any blocking call (std::fs::*, image::open, a
subprocess .wait()) that could otherwise block indefinitely against an
unresponsive mount point.
Polls child for completion, killing it if it hasn’t exited within
timeout. Unlike a raw blocking .wait()/.status(), this guarantees
the caller gets control back within roughly timeout even if the child
itself is stuck on an unresponsive mount point.