Edge cases in overlapping solved benchmarks
When multiple strategies solve exactly the same benchmarks in a timeslice we have no control which one gets selected by LP. Hence, we should have a post-solving optimization phase to pick of those strategies the one that has the best characteristics. Best characteristic could be: least sum-of-solving-time, biggest gap to timeoute, ...
An extreme case is when all available strategies solve exactly the same benchmarks. Then LP is called with an empty problem and returns no strategy. This case could easily be detected and catched by the procedure outlined above.