Ok this is interesting - I made a simple benchmark test, and not only does it demonstrate the problem, it shows the pure performance delta is 100 times worse
The benchmark has a fairly simple setup - 3 tables, A, B, and C, and two triggered auto-enter calc fields. One pulls data from A->C, the other pulls data from A->B->C.
The TOG looks like this:
The results are dramatic:
Begin Test 1 (PSOS Server 20.3.1)
Updated 1000 records in .44 seconds.
Rate = 2268 per second
Begin Test 2 (PSOS Server 20.3.1)
Updated 1000 records in 50.88 seconds.
Rate = 20 per second