You might well surprise yourself. The hardest part of the whole thing is accurately distilling exactly what was causing the failure. And this is also where you the user is the best placed.
Mine had to do with Python Datetimes and timezones and after figuring out how to reliably reproduce it with a minimal case I could step through the stack trace and see what the library was doing (not much, just 'adding' some times together) and could see what might be causing it. (I haven't actually got to the bottom of it yet - I had other stuff to do.) But really the hard and time consuming step is that first bit. Of course, like anything it's a cost benefit - if it'll take weeks to nail it down and fix, vs using something else - then of course, a new approach is needed.
Mine had to do with Python Datetimes and timezones and after figuring out how to reliably reproduce it with a minimal case I could step through the stack trace and see what the library was doing (not much, just 'adding' some times together) and could see what might be causing it. (I haven't actually got to the bottom of it yet - I had other stuff to do.) But really the hard and time consuming step is that first bit. Of course, like anything it's a cost benefit - if it'll take weeks to nail it down and fix, vs using something else - then of course, a new approach is needed.