mishandling of the sacch l1 header
Osmocombb complains about unsupported lapdm sapi=4 messages in the downlink, looking at the wireshark capture it's obvious that this is because it sometimes (but not always?) interprets the l1 sacch header as beginning of the lapdm frame, i.e. si5 in frame 18 and the wrong sapi=4 version of it as seen by osmocombb in frame 21 in the capture.
No data to display