I think that software like mathematica is a great tool, and anyone genuinely doing research that would benefit from it should be happy to pay for it as a matter of course. But I can also see many of the downsides to using closed-source software for research (I myself use some, not mathematica but related) - you don’t know what algorithms are being used, and can’t check yourself that they do what you expect, or fix them if they go wrong. For truly research-level questions, checking that an algorithm produces the correct output may be as hard as writing the algorithm itself…