There's some interesting discussion of it on the linux kernel list, referring to this paper. In short, a guest OS which has access to some hardware could raise interrupts on the host OS and cause the code that handles those interrupts to run. At best (from the defense's standpoint) this could cause strange behavior or crashes as the software tries to handle hardware events that aren't happening. At worst it might exercise a vulnerability and lead to a host compromise (people don't tend to write drivers assuming that the hardware lies to them). It's somewhat of a corner case and lots of things need to go wrong to make it a problem, but there's a warning for it anyway.
If you're not running VMs on the server, don't worry about it. If you are, well, you may want to look into hardware with a newer (smarter) BIOS :)