Problem

Five finally blocks call fcntl.flock(fd, LOCK_UN) without a try/except guard. If the unlock raises OSError (e.g., file descriptor already closed, NFS stale handle, kernel interrupt), the subsequent fd.close() is skipped and the lock file remains open and locked indefinitely.

This is the same pattern that was already guarded for msvcrt (Windows) in every one of these files — but the fcntl (Linux/macOS) path was missed.

Affected modules

Fix

Wrap each fcntl.flock(fd, LOCK_UN) in try/except OSError: pass, matching the defensive pattern already used for msvcrt.unlock in the same blocks.

Before vs After

# BEFORE — if unlock raises, close() is never called
finally:
    fcntl.flock(fd, fcntl.LOCK_UN)   # ← can raise OSError
    fd.close()                        # ← skipped on error

# AFTER — unlock failure doesn't prevent cleanup
finally:
    try:
        fcntl.flock(fd, fcntl.LOCK_UN)
    except OSError:
        pass
    fd.close()                        # ← always runs