I agree with all of your points with the exception of Flickr. I've always thought it would make more sense for a fledgling startup with any investment capital to speak of  to make a run for one of the low earner agencies. Infrastructure, contributor base and at least some customers already in place. Build from there, rebrand and restructure as necessary more along the lines of what you outlined above. At least you are not starting from scratch.

Like GL?

One look at chart on the right and the earnings ratings tells me this is a challenging market to enter to put it mildly ....a slightly polished version of what's already there is not going to cut it. it would need to be a radical departure.

It might be easier to just salary the artists you want and create a full ad agency/design studio that offers a stock catalog as well. That way you can poach the best talent, have a curated collection that fits your market. It does put you on the hook though to create enough business and clients to actually pay your artists and make a profit.

The beauty of the microstock model is that contributors bear a huge proportion of the risk...thats why some agencies drift along apparently selling next to nothing...

And that is the barrier to entry to new agencies too. They don't really offer anything because they don't have any clients or business.