"Burstable" RAM In VPSs

Most VPS offerings have "burstable" memory allocations. I get how you can do this with most things -- letting a VPS access more CPU cycles can be done pretty easily, and a 10 Mbps line burstable to 100 Mbps makes sense.

But how does the guest OS handle all of a sudden having more memory? Since a lot of VPS guests run without modification, how does this work? Does the typical Linux system support dynamic changes in RAM? I can only imagine that dealing with "un-bursting" is even more complex: suppose I have 128MB, burst to 512, but then the host node tries to reclaim some of it. Is the system smart enough to seamlessly swap out to disk when RAM "disappears?" I'd fully expect a kernel panic when the system's RAM shrinks in size, particularly when the RAM was in use.

I'm curious about exactly how all of this is managed. Given that 95% of VPS hosts give a burstable range of RAM, what exactly manages this? Is the virtualization platform handling this and somehow "tricking" the kernel, or is the guest system able to deal with changing RAM allocations? And, if the latter, is this a standard feature of the Linux kernel, or are guest OS mods still necessary to deal with burstable RAM?