I guess we've been fortunate and haven't had to pull logs for Vmware tickets in a while. We had an issue over the weekend where we had to hard boot hosts to resolve. Afterwards, we pulled logs for a vendor to look at but, they only went back to the reboot.
It's been a while but, I seem to recall the syslogs just overwriting based on how much space was available regardless of reboots. Has that changed?
One thing that we have changed is booting our hosts from SD cards rather than local disk or SAN. Could this make a difference?
Yes. Since you are booting from SD cards you do not have persistent storage for your logs. You'd need to forward it off to a syslog such as LogInsight or comparable. Your hosts should be throwing an Informational also:
Is it different if I were booting from HD?
Yes, because when you install your OS on a HDD ESXi will create a separate partition if conditions are met. I believe if the boot device is not SD Card/USB Flash and is greater than 8GB, the difference would be formatted with VMFS as a datastore where your scratch partition and core dumps would go to by default.
Some of this has changed since 7.x was released to accommodate for limitations in system partition and boot bank sizes.
That's the thing though, our SD cards are 32GB. I would have thought that was large enough to put the logs on. But it seems that Vmware sees it's an SD card and doesn't even bother.