I'm going to take a guess that the "high demand for large-scale instances" might gum up the whole thing if it attempts to reserve a server for one of those instances and can't find space.

Suppose the "instance allocator" works strictly on a first-come-first-served basis. There's a new Eureka instance that needs space - and until there IS space, nobody else gets in. Combine that with it maybe not being too smart about "this is too big to actually fit" (not having been designed to handle such large groups) and it might be waiting until enough other instances complete and close down to make that space available.

Meanwhile, everything else that should have gone through is queuing up behind it.

It's just hypothetical, but it might explain why not even smaller groups and solo instances can get in.