Wow.  Looks like ref counting is a winner.  Simple local mode test, comparing 5.2.0.Beta2 (in red), my fix for ISPN-2381 *without* ref counting (in blue) and with ref counting (in green).

Weird that read performance is also affected, as these patches are for the lock manager which is not used when reading.  Could be overall resource contention that slows down reader threads.