Yes that's correct - when I resume after a period of many hours, e.g. greater than 12 hours for example, the system seems to perform a full boot with the POST startup (beeps) etc. As such, the contents of RAM as they were at the point of the suspend, are lost.
I should also mention that the server OS is Fedora 20 Linux, with the latest kernel update at the time of writing (3.17.4-200.fc20.x86_64)
Note: I thought I had found a workaround by disabling the "Reboot on CATERR" option in the BIOS, which seems to prevent the issue, but occasionally seems to cause a different problem, whereby the server reboots immediately after initiating a suspend (as in, the system suspends, but then immediately resumes into a full boot with POST etc.)
The fact the situation only takes place when and only when "[resuming] after a period of many hours, e.g. greater than 12 hours" (qtd. before), not only does it prove ACPI S3 does work, but it also shows the cause is something else. In light of this, and since Operating System-wise, you confirm it has the latest kernel updates, I would recommend the following (see steps below) in the event there is a BIOS corruption:
1. Clear the System Event Log and load up the BIOS default settings. For this, please, proceed as follows:
1.1. Boot into BIOS.
1.2. Once in BIOS, hit the F9 key to load up the BIOS default settings.
1.3. Go into the Server Management tab and enable the "Clear System Event Log" option.
Note: Reset any customized BIOS settings before continuing.
1.4. Hit the F10 to save the changes and exit.
2. Test. If the situation still takes place, because the most current BIOS one version is 02.02.0005, please, update this (click here). The instructions on how to run the BIOS/firmware updating are found here:
Please, repeat step 1.2 above right after the BIOS/firmware is updated.