First, I have no idea if this is related to the leaning problem, but using a bad STL file isn't a good thing to use to identify the root cause. As I mentioned in a previous post, I have seen this problem but not often. Unless something can be duplicated I am hesitant to point any fingers at either S3D or Leapfrog.
Second, S3D should ideally flag a bad STL file as soon you load it (Kisslicer does). No reason it couldn't run the manifold test automatically.
As far as repairing the bad STL file, S3D does have some built in repair options. Depending on how bad the errors are that may or may not work. There is a utility called Netfabb (the basic version is free) that can sometimes fix the bad models. There is another utility called MeshLab that can too, but there is a pretty steep learning curve.
I find that quite a few models that I've downloaded from Thingiverse.com have errors.