Some things to think of:
1 - it must be hard to generate one block, but cheap to verify. I doubt that the simulations in this context can be easily verified.
2 - difficulty must be adjustable to keep block rate mostly constant. Of course, you could adjust the simulated time for each unit, don't know how well that would work.
3 - each block must depend on the previous block, i.e. it must not be possible to perform all or part of the hard work before the previous block is known.
I have a gut feeling that most grid computing efforts don't fit these constraints well.
Onkel Paul
2 and 3 may be manageable through centralization but then you're out of the bitcoin "family" and into another mess of issues. The need for one way functions- point 1- is where you would really need to work some algorithmic magic to make many types of problems workable if they could be made to work at all.