There seems to be no reason to call sr_dev_close inside of std_serial_dev_acquisition_stop. This actually breaks pulseview because multiple run/stop cycle don't work. This hasn't been an issue so far because pulseview supports multiple acquisition cycles and it only shows logic analyzers / scopes and none of the drivers for these use std_serial_dev_acquisition_stop. The only serial based LA is ols and it has its own acquisition_stop implementation. The only drivers I could find that use std_serial_dev_acquisition_stop are DMM/LCR and those work fine when used in sigrok-cli because there is only one open/start/top/close cycle in that case and so the fact acq_stop calls dev_close doesn't prevent things from working.
Created attachment 470 [details] Proposed fix
Fixed in cfdc80151b106ae83f3380cb308d53314974257a (added bug number to the commit msg), thanks!