Exactly this. The only pile that might feasibly be reduced to zero, given a sufficient number of IDers, is verifiable observations without an ID (“unknowns”), because it is always possible to either add an ID (even that of “life”) or make them casual. For everything else, whether at kingdom or class or even lower levels, there will always be some percentage of observations that IDers do not manage to resolve because the photo quality is lacking or they are tricky for other reasons, and these will tend to become more over time. If we accept this and adjust our expectations accordingly, the task of IDing becomes much less disouraging and frustrating than if we approach it with a belief that it should be possible to reduce these piles to zero.
There are lots of legitimate reasons why people don’t always mark observations as “ID cannot be improved” (including not wanting to make observations casual just because the ID is above family level), though I think there would probably be benefits if experts would use it more often for observations at genus or lower when they know that there is not enough evidence to determine the species.
We’ve had this discussion before.
There is a nontrivial problem of how to determine whether someone has “seen an observation and moved on”: Does just looking at thumbnails in Explore or Identify count or do they have to mark it as reviewed? What about people who go through a lot of observations to annotate them but are not trying to ID them?
Beyond this, the fact that someone has seen an observation and maybe even marked it as reviewed says very little about whether the observation is in fact identifiable to someone with the relevant expertise.
Some people use the “mark reviewed” function as a way of removing the observations of taxa they are not interested in from their Identify searches. Specialist users may skim observations that have a broad ID in order to pull out the ones that are relevant to them – e.g., I often browse broadly ID’d hymenoptera to find the bees and aculeate wasps, but there are numerous hymenopteran families I simply don’t know enough about to narrow the ID. A reasonable portion of these observations are probably identifiable, but I am not the right person to do so. It would be a mistake to assume that because I have seen an observation but not added an ID, it means anything more than, say, that ants all look pretty much alike to me or that I haven’t spent any time learning about sawfly larvae.