I might just add to this that there are several (lots) of scientific compute problems that are trivial to code relative to the amount of compute they can consume.
Some examples:
The search for primes can consume an infinitely amount of compute, forever, until someone develops an algorithm to find all primes. This may never happen.
The interactions of molecules. On top of just folding individual proteins and 1-1 interactions, there are endlessly complex poorly understood reactions that can be investigated. If we had infinite compute, we could as an example take a virus and model the interaction of every known molecule with it.
Analysis of data from space-ward arrays is limited by analysis throughput. If we could guarantee 100 TFLOPS of throughput, many institutions and arrays would be more than happy to gather enough data to fill that processing bandwidth.
Thanks! These are great examples. I touched on a couple and will expand them. I thought about mentioning primes and will do so now!