PVE 9: More issues with pvestatd. CPU graphs spiking to 200%?

[dw-cj]

Since upgrading to PVE 9, I've noticed VM CPU graphs spiking to crazy numbers, as high as 200%+. On PVE 8, I never saw anything higher than 105-110%, and even that was rare. I understand that NIC / disk emulation can cause for CPU to go higher than 100%, but these VMs don't really have any unusually high I/O or network usage. I'm seeing this across many different VMs, almost daily. I often see CPU spikes of 150%+, and in extreme cases, even 200%+. Any idea whats going on here? These are QEMU VMs ranging from 4 to 12 cores.

Examples:

 

[dw-cj]

Has anyone else ran into this?

 

[fiona]

Hi,
the accounting in Proxmox VE 9 is much more granular, so it might be that the spikes were already present in Proxmox VE 8, but averaged out. And they really seem quite "spiky". Are there any special operations (e.g. backup) being done during the time of the spikes?