Here're advantages:
- thick/unison FM8 sounds could take up to 40% CPU time here on my system, and then you've got 1 tone only, sample it, and it'll become one intrinsic voice again, cpu-wise
- otherwise densely layered sounds (multiple instruments, multiple notes) become one
- easier to make 'drumkits' or otherwise multiple instruments (with limited key range) per sampler track
Some disadvantages:
- You may loose some of the means to control your sound, for instance sound-wise changes and tiny timbre variations
- it's a h**l of a lot of work!
For me, if the sound isn't too radical in changing timbres and I see an opportunity to save CPU cycles, I usually sample FM8 sounds. Last night I made an analogue'esque FM8 horn using 20 intervals (each 3 semitones) and 16 velocity layers. Took hours, but in the end I gain a very playable instrument (of over 200MB on HD
