It was not advertised well but the HEPI isolation loop actually caused the alignment jump of [+25, -8] urad according to IPS in HEPI when it was turned on Friday evening (attached). We've never had a good alignment after that but this might be a coincidence. This is not a huge deal but annoying nevertheless.
When I started to restore the known good alignment of HEPI by giving new setpoints to HEPI it tripped the watchdog repeatedly because IPS completely railed, and later I was told that that was the reason why SEI gave up on restoring the original alignment.
Anyway, I was sufficiently annoyed and relieved the HEPI offset by pushing TMSX as far as we can in a hope that this is the last time I'm annoyed by the difference between offset-only and isolated (because we'll always isolate from here on).
In a retrospect I didn't have to go back to offset-only, I could have just changed the setpoints, but that's what I did due to various reasons. Anyway, In the end, I obtained this: (H1:HPI-ETMX_IPS_RY_SETPOINT_NOW, RZ)=(-389800, 111650).
I also copied these numbers and put them in H1:HPI-ETMX_IPS_RY_TARGET and RZ.
IPS are still yellow (about 20000 cts on V1 and V2) but not railing.
Just so that it's clear: