Gail Badner I've found out that it causes many putFromLoads to fail, so it's not an optimization but it prevents harmful effect I haven't realized in the past. So I'd say 75% bug 25% enhancement.

Btw., I've found out when running performance tests, but I am still puzzled seeing too few entries being cached in nonstrict mode vs. read-write mode (both on repl caches) - while I have this locally fixed, I am looking for yet another deficiency.