From a pharmacology/activity point of view much of organic chem can live with categorizing at the connectivity level (which is part of why SMILES works most of the time). But some of it really does need to dig deeper to the stereoisomers and polymorphs. This doesn’t mean everything in the db has to go down that far in detail but that the db needs to allow for the cases that need it. So the database ought to allow a general carbon record which can point to additional records, one for each of the polymorphs.