>> It wouldn't have been 9689, it doesn't touch dvbsiparser. And I didn't
>> see any mismatched locks in there, but I applied the patch anyway since
>> using a QMutexLocker is much cleaner than raw lock/unlock calls.
> The mismatch was in DelPid (we can return early without releasing the lock
> halfway through). In 9689 we call DelPid much more than we did previously,
> increasing the likelyhood of the problem.
I'm now running SVN 9710 and it seems to work, no hangs anymore.