are there additional key learnings that occur as part of the back-and-forth interaction between manufacturing engineers and design engineers that couldn't be obtained from a pure reverse engineering of the silicon itself?
That's hard to say. It would depend on how the foundry operation was set up an run. I suspect that Intel would only offer a limited subset of its processes (keeping its high performance MPU crown jewels stuff off limits) to outsiders and would hire new customer facing engineers for its foundry operation who had never worked on their own design teams. The only process specific IP flow would be from the process guys for things like ESD protection structures.