Hurrah!  Finally, a (partial) resolution to a problem that's been around for over 20 years...  Yes, a full error set would be nice, but I suspect that that would involve a serious overhaul of the error handling system.  But this is a good step forward.  Thanks, Steve, for bringing it to our attention.

...but did you say it's TF-dependent?  Why?  Why can't they just surface this extra information by default?  Is it because they're putting it out as a different error message ID?