A comprehensive creepage/clearance engine would be cool, yes. You'd need ways to input boundary conditions for 3D models, to demonstrate issues like what are pictured. That would take a lot of work, and I don't think STEP files are very suitable for that (even when they come in proper colors rather than solid uncolored mesh..).
More work just means more opportunity to screw up, and more places to hide those screw-ups. Consider the audit output of such a check: it might include the rule settings (so you can see if any quirky things have been special-cased), and the results of applying those rules. What's likely not in that audit, are the data which went into the check: boundary conditions of the components, which conducting and insulating features resulted in those decisions, and what distances and paths those objects passed or failed on. When the shit hits the fan and a cause (or scapegoat) must be found, all these data will be brought to light, sooner or later, and at much greater expense (e.g., discovery in a lawsuit) than checking it out in the first place.
The present "best practice" is to subject the design to an independent review, i.e., UL or other NRTL. They are just as fallible as any other engineering firm, but they back that up with a guarantee: they take liability for design failures, not you. It's expensive, but not much more so than hiring a consultant to check the work -- and that wouldn't carry a guarantee.
So the key there is: work. If it can be done with less user input, more automation and greater reliability, it's a win. Otherwise, it's not much better than how things are today. Sucks, but that's how things go.
Tim