Hi,
FYI it was an ultimate example why one should not perform full-blown config for X440-G2s in a stack that exceeds best practice size. The stack that made issues was 7-slot, others were 2-4 slots only. Because of that most likely, the master's CPU was heavily loaded, reaching 100% quite often. Probably that's why we've seen many logs about mirroring instances, and also ports started to flap (at least those with APs connected). Huge mess. š
Disabling telemetry left the switch free of high CPU utilization and AP link outages, next step is to simply split the stack.
Hope that helps,
Tomasz