Would be difficult to say what's causing the lockups without the logs for the time frame. If the crashes are happening with the untimited CPU option selected, could be a CPU contention for all I can think of.
Say those Debian machines, how many vCPU's are they given? And how much CPU resources are they consuming? But more that that, what exactly are the VM's being used for?
Also what controllers are they using - the VM's I mean?