The fact that this change also makes observations casual if the observation ID is not the same as the community ID when the DQA “ID cannot be improved” is used is also anything but ideal. These observations are not broken; the only “problem” is a mismatch between observation ID and community ID. They should not be casual.
It means that if a user uses the DQA before there is a community consensus – or if this is lost due to a user withdrawing an ID – and doesn’t notice that their action has made the observation casual, a perfectly good observation will essentially end up being thrown out instead of being returned to “needs ID”, because very few people look at casual observations. This affects observations, for example, that have one or more genus IDs and a single active ID at some level between genus and species (subgenus, section etc) and the DQA has been used to take it out of “needs ID”.
It seems that what we need is for the DQA to only be active (both in terms of being votable and in terms of how it affects the observation) as long as the community ID is the same as the observation ID. But note that this would leave us without a way to deal with opted-out observations that the community has tried in vain to refine, so maybe there would need to be an exception (I think this would be covered by something like “community ID must be the same as or more specific than the observation ID” but I would need to work through more possible scenarios).