It seems like it's tricky to inform a threadpool implementation (like programming assignment two) of domain knowledge that could help wish such scheduling. In particular, the interface abstraction makes it opaque to the pool and so it appears that all tasks are the same to it. Perhaps the interface could be amended with a function that gives some estimate of the difficulty of the task to the threadpool.

I think it is very hard to estimate the difficulty of a task since its execution time could depend on a lot of things such as scheduling and we can not know what the future is going to be like. So I think it's quite important the programmer understands dynamic and static scheduling and know when to use either of them.