This is a good question. I can assure you that my use case is so special that no-one else will try to do this... :)
I am a PhD student doing research on recommendation systems for software engineering. More specifically, I'm building a platform that eases the research on such systems, by -for example- providing preprocessed datasets and labelled data that can be used as the ground truth for evaluations of the machine learning part (see [1] for details).
In the course of my studies, I created two datasets: A dataset of interactions of developers within VS that I collected with our KaVE plugin [2] and a second dataset that contains preprocessed versions of many C# solutions from GitHub: the preprocessing captures a simplified representation of the AST for all contained types in the solutions. As an extension, I would also like to preserve some information about the type system in a simplified form. This is where this thread comes into place...
Technically speaking, we are "abusing" the BaseTestWithExistingSolution to open the solutions in the cloned repos. We are iterating over all contained types and transform the AST (and soon the typing info :)). Even though it feels that this base test was never designed for such a task (don't ask me for a list of bugs :)), it is the only way we found to bulk process that many solutions, after the plugin support from the R# command line tools was dropped (with the transition from R#8.2 to R#9.0, I think).
Does this answer you question? :D
best
Sebastian