I'm thinking about a small vSAN stretched cluster in two different sites + a third for the witness.
From a availability and resiliency stand point, are there some advantages using a stretched config of 2+2+1 instead of 1+1+1 ?
If I'm not wrong both have PFTT=1 and SFTT=0.
If the storage usage was sufficiently low (or even just the FTT=1 data usage being relatively low and FTT=0 for other data) then a 2+2+1 would have the capability to rebuild all data back to FTT=1 with one node failed on one site, whereas a 1+1+1 cluster with one failed node would be FTT=0 until the failed node was functional again or replaced.
I meant if you had a mix of FTT=1 and FTT=0 data - FTT=0 data won't ever be resynced anywhere because it either is wholly available or it is not (and thus it will never require any more space than it is currently using other than if it is ever being relocated).