US Government has used selective availability from time to time - this introduces a deliberate error into civilian GPS recievers and GLONASS being Russian would be no different. Most Survey equipment, even RTK gear takes several minutes if not longer to initialise and UAV users are turning on devices and flying away in minutes if not sooner.
As an engineer I would like to see some testing on the minimum number of sat's required for a good fix (not just the DJI >6 statement for an A2 or wookong system) to determine if the sensor has had sufficient time to really know where it is AND if the flyaways do not tend to be in the direction of the previous flight - possibly some coordinate artefact in the system somewhere, you loose signal and it heads to a known coordinate ??