Comment 2 for bug 1627749

It's a tricky situation not fully resolved in ml2 or anywhere else.

What if a driver failed, but some other succeeded ?

Shall we revert them all, and then revert the DB? what if one of the reverts then fail?

I think that we should let the drivers report an error via an exception (without making other drivers fail), but at the same time, it should be the failed driver responsibility to track the need to resync such object when the backend comes back (otherwise the revert failure becomes an issue too).