Why don't you do an eboostr speed test with one, two and three drives. You could also do a read test on all the drives at once to test the USB2 bandwidth (not sure what software does this).
I don't know how eboostr works, but at best eboostr collects stats and is load balancing using the fastest drive first, at worst it is querying all drives and using the first result... so in all cases you will want the cache on the USB's to contain identical copies of everything not in ram.
If the USB controllers are not sharing bandwidth and assuming the thumb drives are 30-32MB/s already, then pure theory says you may get 60MB/s from two fast drives, and 90MB/s from 3 under heavy load (windows or the controller may just cap you at 30 or 60 though). It might also get past the System Reserved 10% bandwidth for one drive.
There may be other side-effects of using lots of drives (like an unresponsive mouse), it probably depends on what sort of system you have.