Still not sure how this would work. I will try to test on multiwallets this weeks, but I have still no idea how this could work. Shouldn't be a multiwallet miner even be slower because of the overhead of multiple instances running?
Well what I was thinking is 40 threads, on 24 wallets-per-server-instances will theoretically be hashing similar block timestamps with the same mining transactions (think of when we have 1 tx in the block). With a 40*24 set (880 threads) will be increasing the low nonces slowly. There may be a propensity for more than one thread to arrive at a legal pool solution almost at the same time, whereas a single miner, mining with a couple PCs and 10 threads will be more likely to solve a distinct low-work share (in contrast to more than one share).
I'll pm you here and maybe we can create a test plan.