After a small tweak, I was able to get my idea of RIMS+loop to work basically as I envisioned it in a water test. A ball valve controlling the loop flow rate more linearly (compared to the butterfly valve used in my first revision) proved to be essential.
