One problem this does is it shifts the costs of this to individual projects.

In my field, the average OA field is probably around $3000.

Assuming a five year project that produces 5 papers a year (not unreasonable), you get $75,000.

That's almost exactly what it costs, with salary, tuition, and fringe, to support two graduate students. So funders would have to accept getting less productivity from the same projects.

It also favors large labs with senior PIs that have the funding to absorb those costs.