...The Corsair P128 does use MLC but has a much improved controller (and therefore as good or better performance than SLC) -
Sales an marketing will say anything to make their product look good. 
...I know I'm in slightly unchartered territories with SSD's and VM's.
If you already have the drive, or are definitely going to get one, then I'd just try it. I got my SSD because I was turing a older Sony laptop into as much of a "netbook" as possible. So the performance of the SSD I got was not a critical factor as much as the power consumption. But I tried it just for kicks because like you've seen, there aren't that many people with SSDs yet. So you're on the bleeding edge and most likely your results will be the reference for the rest of us. 
Sales an marketing will say anything to make their product look good.
I possibly stretched it a bit by saying this particular Corsair drive is as good as SLC. But if you look at the real world and hypothetical benchmarks (from many different review sites), this drive is vastly superior to older-generation MLC drives (and blows away 10K RPM drives) and is almost as good as the Intel X25 (SLC-based) series.
My original question still stands though: What can I do to maximize performance and minimize degradation of the SSD for use with a Virtual Machine? From what I gather, I should go with a preallocated disk to minimize disk writes? What about splitting into 2GB chunks? I also came across some recommendations for (real, non-virtualized) XP machines to disable to following via regsitry:
Disable Automatic Boot Disk Optimization
Last Access Time Stamps
Enable write caching on hard disks
I have no clue if these are even used in a virtual machine, though. I certainly can go the trial-and-error route but would appreciate any insight or advice.
Would be interested to know how you got on as I am looking into this too. I came across this thread while searching for more information on the topic, but I already had some ideas about how to do this. Im thinking of VMWare workstation here, but I think the principles are identical although I confess my fusion-specific knowledge is essentially zero.
By using a linked clone on the HDD from a template on SSD, you get most reads from SSD and most writes to HDD (at least until linked clone starts getting very large), which is optimal. Of course there is a performance hit using a linked clone but I suspect with the fast read speeds of SSD, the hit in real world situations would be negligible.
Once the clone becomes too large to see these benefits, just commit (by creating a full clone of the linked clone and overwriting template) or create a fresh linked clone as required.
You could also create separate virtual disks for commonly read and commonly written files as are you are encouraged to do on your SSD equiped host.
Im not sure if there is anything like MFT (which seems almost a requirement in Windows to prevent stuttering and accelerated burn in environments with frequent small random writes that I would expect VMs to be a prime example of) for Mac, but if so that would be a good idea too.
Of course, all of the above is pure conjecture until I actually get some disks and do some benchmarking