Is your system overcommitted? Have you grant your VMs more memory than you really have physically? I have seen some ESXs starting to swap even if the VMs were not using their memory but just because we gave them more than 140% of the physical capacity of the hosts.
Do you see any other signs that the host is constrained on memory? Do you see any other ballooning or memory compression numbers?
If not then maybe it was just a situation where VMs were rebooted at the same time and the host temporarily ran out of memory. And I realize this is a stupid question but the swap values you're referencing there are on the ESX host and not inside th guest, correct?
Correct - this was vmkernel swapping, not internal OS swapping.
I know that there was a reboot of the system and all VMs at about 5PM that day but the brief spike in swap activity appeared about 1 hour previous to that.
There was some balloon activity at the time but that was due to the customer setting a limit on memory of 16GB when the configured memory was 32GB. SQL was running and this scenario causes ballooning to occurr. But the system as a whole was not under memory constraint. In fact, I would say that it was only up to 50% usage at the time.
No, usage was at max about 50% of the physical memory.