Just to update, still getting problems with RPC connection and BDM errors using SDM mode on 93.0.82. The BDM errors are reduced to a rare event, but still occasionally happen (1 so far since testing .82).
I'm fairly sure all/most of these BDM are false positives, as the BDM reports zero problems on subsequent runs whether you re-build Db or not. It'd be good to hear ATI's opinion on that. I'm fairly certain that the old Armory Db would fail deterministically once an error was detected.
I think we're not gonna touch the RPC comm layer any furter for now. It needs to be reworked to a certain extend and that's too disruptive for the class of fixes we are aiming to ship in this version. I personally think the remaining RPC errors aren't deal breakers, and that our users will have the good grace to endure them until I come up with a rework of that 2 years old code (instead of band aid patches here and there, which haven't played up to expectation).
As for the BDM errors, I dialed down the error message because I expect all left over issues are false positives. The BDM error just shuts down Armory now (rebuild and rescan isn't checked by default in the error message), and the next start will just repair the issue (if there is one indeed) until I get to the bottom of it.
I did push a tiny but significant change a couple days ago which should fix some latent BDM errors.