I did some digging and it appears that the change up to 72 in 1fb5395 is
before the 2.96 release.
Also, this code in driver_nmea0183.c gives the warning:
if (session->driver.nmea.part == session->driver.nmea.await
&& atoi(field[3]) != session->gpsdata.satellites_visible)
gpsd_report(session->context->debug, LOG_WARN,
"GPGSV field 3 value of %d != actual count %d\n",
atoi(field[3]), session->gpsdata.satellites_visible);
So maybe we have a bug in the firmware of our gps devices. We'll contact
the vendor of those.