How to reproduce:
1) Start PV
2) Load a trace, e.g. sigrok-dumps/uart/hello-world/8n1/hello_world_8n1_1200.sr
3) Add a decoder, e.g. guess_bitrate or timing
4) Assign the channel to the decoder
5) If the decode doesn't crash PV, keep clicking Run after the previous decode finished

Expected result:
PV never crashes

Actual result:

srd: Created session 5.
srd: Creating new timing instance.
srd: Created new timing instance with ID timing-1.
srd: Setting channels for instance timing-1 with list of 1 channels.
srd: Setting channel mapping: data (index 0) = channel 0.
srd: Final channel map:
srd:  - index 0 = channel 0 (required)
srd: Setting session 5 samplerate to 625000.
srd: Registering new callback for output type 0.
srd: Calling start() on all instances in session 5.
srd: Calling start() method on protocol decoder instance timing-1.
srd: Instance timing-1 creating new output type 0 for timing-1.
srd: Initial pins: 0.
srd: Decoding: abs start sample 0, abs end sample 4096 (4096 samples, 4096 bytes, unitsize = 1), instance timing-1.
srd: No worker thread for this decoder stack exists yet, creating one: timing-1.
srd: NULL/empty condition list, automatic match.
srd: Instance timing-1 put 0-389 OUTPUT_ANN on oid 0.
srd: SystemError: Protocol decoder instance timing-1: : <built-in method wait of Decoder object at 0xb550718c> returned a result with an error set
srd: Instance timing-1 put 389-2472 OUTPUT_ANN on oid 0.
srd: Instance timing-1 put 389-2472 OUTPUT_ANN on oid 0.
srd: StopIteration: Protocol decoder instance timing-1: : 
QObject::killTimers: timers cannot be stopped from another thread
srd: ['Traceback (most recent call last):\n', '  File "/home/abraxa/repositories/sigrok/build/share/libsigrokdecode/decoders/timing/pd.py", line 111, in decode\n    self.last_n.append(t)\n', 'StopIteration\n']


I've also encountered this kind of error for timing:

srd: No worker thread for this decoder stack exists yet, creating one: timing-1.
srd: NULL/empty condition list, automatic match.
srd: SystemError: Protocol decoder instance timing-1: : PyEval_EvalFrameEx returned NULL without setting an error
srd: Instance timing-1 put 0-389 OUTPUT_ANN on oid 0.
srd: Instance timing-1 put 389-2472 OUTPUT_ANN on oid 0.
srd: Instance timing-1 put 389-2472 OUTPUT_ANN on oid 0.
srd: Instance timing-1 put 2472-2993 OUTPUT_ANN on oid 0.
srd: Instance timing-1 put 2472-2993 OUTPUT_ANN on oid 0.
QObject::killTimers: timers cannot be stopped from another thread

...and this one for guess bitrate:

srd: Created session 18.
srd: Creating new guess_bitrate instance.
std: virtual-session: Sending SR_DF_END packet.
sr: [00:24.530062] session: bus: Received SR_DF_END packet.
srd: Created new guess_bitrate instance with ID guess_bitrate-1.
sr: [00:24.530143] session: fd_source_finalize: key 0xffffffff
srd: Setting channels for instance guess_bitrate-1 with list of 1 channels.
sr: [00:24.530291] session: Stopped.
srd: Setting channel mapping: data (index 0) = channel 0.
srd: Final channel map:
srd:  - index 0 = channel 0 (required)
srd: Setting session 18 samplerate to 625000.
srd: Registering new callback for output type 0.
srd: Calling start() on all instances in session 18.
srd: Calling start() method on protocol decoder instance guess_bitrate-1.
srd: Instance guess_bitrate-1 creating new output type 0 for guess_bitrate-1.
srd: Initial pins: 1.
srd: Decoding: abs start sample 0, abs end sample 4096 (4096 samples, 4096 bytes, unitsize = 1), instance guess_bitrate-1.
srd: No worker thread for this decoder stack exists yet, creating one: guess_bitrate-1.
srd: NULL/empty condition list, automatic match.
srd: ZeroDivisionError: Protocol decoder instance guess_bitrate-1: : float division by zero


The fact that none of these issues appear immediately but only after several runs tell me that there's something broken generally. Now if only I knew what :)