In other words: you don't get a stack trace, so it's hard to debug. It's like writing one big function so that your stack traces tend to be very shallow, and then the only clue to what's going on is the _line number_ where something goes wrong. But that's not so bad, is it? You get to know where in the processing of something you were, and that's a pretty good clue. You'll want to have an option to emit traces, sure -- that will be a lot like a stack trace, but better. And you might want to have traces held on to speculatively and then throw them away at select points (like when you're done processing a request w/o exceptions).