I looked at this, and it looks like your data is too clean to be representative, because your sets are disjoint and your subsets are complete.  That is, if you define a set A as every test that can be run in parallel with test a, then every test within that set can be run in parallel with any other test in that set and cannot be run in parallel with any test that is not within that set.

In the real world, I would expect things to be much messier.  For example, say tests A and B use the same set of equipment, so they cannot be run in parallel, but test C uses a different set of equipment, so it can be run in parallel with either or both A and B.

In this case,

Set A = {A, C}
Set B = {B, C}
Set C = {A, B, C}

Sets A and B are not disjoint, because they are not the same, but they both include {C}.  Set C is not complete, because it includes tests A and B which cannot be run in parallel with each other.

I came up with a solution that works only if the sets are disjoint and the subsets are complete.  I haven't posted it, because I don't think that it will work for your date, because I don't think your data will match that criteria.

Drew