lolroger
Yes, we are going to likely ship an "Other models" bin in the picker for the more "dubious but possibly still relevant for some people" category.
Generally we would still prune models that fall too far out of use / behind in quality, just to keep a lid on UX complexity. We try not to think of the average user of someone who is "in the loop" enough to not be overwhelmed by choices, or get into patterns of using worse models for no particular reason than lack of ecosystem understanding / by accident, then cancelling because it's bad, ...
Put another way, our strategy is to offer everything that is considered "best in its class" for some particular vector, and not just "everything that we could possibly offer just because we can". Keeps the surface area of problems small (UX, provider availability, tool compatibility, ...), we don't have to design around having 100s of models available to the user, ...
