water may not be important for non-polar/hydrophobic molecules but non-polar solvents like benzene maybe used. Generally reactions performed in a solution.
When the molecule structure exhibits complex electronic effects there should be carried out computations of the solvent effects together with the gas-phase calculations.
Solvent effects can be as important for non-polar molecules. In this case, however, its not the "classic" electrostatic solvent effects, but dispersive interactions. It a common misconception that the latter are much weaker an can be neglected. I think this misconception is at least partially rooted in the fact that the most common approach to solvation (PCMs) do can not describe dispersive interactions. For this, one has to use more complete and less well known models, like e.g. CMIRS:
Article CMIRS Solvation Model for Methanol: Parametrization, Testing...
Apart from all of that, it very much depends on what property of a molecule one is interested in. For optical excitations, there is a very similar situation and its often assumed only polar solvents exert a large influence, e.g. on polar charge-transfer states. However, these are also strongly stabilized in non-polar solvents (e.g. hexane) compared to gas phase, which is not very well known. See e.g. this work for a couple of examples and a discussion thereof.