If a molecule solves a particular problem, then it is neither so unstable that a poor simulation could predict its instability, nor so stable that a poor simula

If a molecule solves a particular problem, then it is neither so unstable that a poor simulation could predict its instability, nor so stable that a poor simulation could predict its stability. The molecules that are the easiest to predict computationally tend to be the least interesting. The energy difference between the bound and unbound states has to be small for it to be interesting which automatically puts it in the realm of difficult simulation. So no, chemists do not pick difficult molecules on purpose - they do so because difficulty and interestingness originate from the same place.