Aha, yes, I had a fuzzy memory of such a thing. Unfortunately, I think that means using the fully qualified name isn't a good option. 260 goes really fast if a user organizes their files in subfolders. 
I can't help but wonder whether odkmeta and other downstream tools could use the contents of each CSV to determine what it represents rather than parsing the filename? It seems that might be less brittle.
Either way, at this point I'm leaning towards merging https://github.com/opendatakit/briefcase/pull/659 for now to revert the FQN behavior introduced in 1.12.1 (sorry I got that wrong above, @ggalmazor) and continuing the conversation.