Steve,
thanks a lot for the further hints. Sorry for replying delayed (I have too many urgent things here at the same time at the moment, taking care of the vmware hosts is not my main function)...
I will check the BIOS settings at the next opportunity, when I get a chance to reboot the hosts without causing too much friction. Luckily the issue doesn't happen that frequently. But I am curious to check if there are any power save options set in the BIOS.
Thanks also for the hints on the Windows pausing issue. This just happened again yesterday and I saw in the vsphere client performance plot that the CPU load hit full load at that moment, then it seems things got very quiet on the host for about a minute (maybe 5-10% total load), then things went back to normal. The "quiet period" was about the time when the server no longer responded (to RDP, SMB, web services running on there, tasks running on there seemingly paused). It was only for maybe a minute though, not as long as it has happened before. The host had only been at 100% load for a short time, maybe a minute. The host is running with local disk only (4 resp. 6 SAS 15k drives in RAID 1).
I have also looked into esxtop now comparing to tresholds recommendations on the web. It looks like in normal operation for lower priority and less active VMs there is some memory swapping and ballooning happening - but not for the high priority VM I had the issues with yesterday. Memory status is "high state". For CPU I can see occasional short moments where %RDY jumps up, mainly for lower prio VMs, most of the time it is in the range of 0.x to 3% for all VMs though. Disk load seems fine. I will look into recording this for a while. I should also note that I took one VM off the host yesterday.
Thanks and regards
Felix