So, I run my test against purepool, but I couldn't verify what happened on the main pool.
Maybe my tests are wrong, but in my tests, a single wallet gets around 25% more shares than a multiwallet setup. Looks like the overhead kills hit here.

My test system: I created an VM with ubuntu 18.04 and biblepay from the official repo. I created a new wallet and run it for 1 hour against purepool with 16 threads on my cpu (with 16 cores).
After that, I created 8 new wallets (each from scratch) and pointed them to a new biblepay adress in purepool, each with two threads. The cpu usage was the same with both setups.
I even gave the multiwallet setup two minutes more time, as they needed more time to fully start.

I tried it with more wallets (16), but my internet connection died trying it. Not sure why, I suspect my internet box died with the many open connections.
Maybe it wasn't enough wallets, but I can't test with more until the end of the week with more wallets.

Was my setup wrong?