You can't fathom how 'top human engineers' could take advantage of exclusive access to a frontier model at ultrafast speeds and limitless token budgets to conduct research?

I have all that and LLMs invariably produce garbage, so, no.

Not every problem is best addressed by a top engineer.

Plenty of the tasks that keep a company running can benefit from good-enough (and better than the competition).

Yes, and none of those tasks require even the current SOTA models.