separate script_path encourages tests to miss common scenarios
scripttest always invokes executables with fully-qualified paths, and can end up selecting an executable invoked without qualification that can't be found via PATH. Aside from being surprising (and undocumented?), it is impossible to test the normal usage. Yes, it's important to hit the corner cases, but it's important to make sure testing covers the common cases well. I think you should leave the qualification/no-qualification thing up to the user, who can figure it out.