For the sake of the internet (and past me, who found this link 48 hours ago when trying to diagnose the same issue), I thought I would share my own experience.
I had similar issues with a new install of PVE 8.2.4 onto new hardware (N305 router). It was *almost* 100% stable, but partly crashed overnight with a GPF, leaving all of the VMs and containers running (including OPNSense VM), but the management console unresponsive. After several attempts to regain controlled, I ended up having to perform a "cycle BRS" to bring it back. ("big red switch")
To cut a long story short, the problem was the RAM. It was properly sourced from Crucial and the correct spec (DDR5 4800), but when I ran an extended memtest86 run on it, it finally returned a handful of errors after perhaps an hour of running. Same with memtester within PVE - no faults initially, but a handful after it had been running for a while.
RAM has been swapped to Crucial DDR5 5600 and all memory tests are now running clear and I'll be very surprised indeed if this does not fix my mild instability issues.
tl;dr: The problem is *always* DNS. Apart from when it's bad RAM.
