try:
        self.lock.acquire()
        ...
    finally:
        self.lock.release()

into:

    self.lock.acquire()
    try:
        ...
    finally:
        self.lock.release()

The old version would mistakenly try to release the lock if there was
error in acquiring it.