Peter pointed out to me that the L2_LOCK filter bank for ETMX had FM8 (EvanR) engaged for both pitch and yaw, while for the other 3 test masses it was FM3 (EvanBR). All 4 test masses should be the same, so I've fixed ETMX and accepted it in SDF so that it'll come back correctly upon reboots. It looks like we've had the wrong filters engaged for ETMX since July 5th. I didn't check the ITMs, but ETMY has been correct since before July 5th.
Recall that we have the option between roll-only or both bounce and roll notches so that we could try running with the roll-only filters, which would give us more phase in the ASC loops so that we could make harder cutoffs. However, we've pretty consistently found that we ring up the bounce modes if we don't have the notches on, so we (should) have been using the combo EvanBR filters for all test masses.
Also, for some reason the decimation filters for the OUT16s on all of the ETMX L2 LOCK filter banks were off, and have been off for more than 100 days (that's how long I looked back for the bounce / roll filters). Evan, Peter and I agree that there's no reason for them to be off, so we've turned them back on.