A lot of the decisions made in designing web.py were also carried over to tornado's web framework (which not coincidentally is also a web.py).

At the time (and I have to admit my memory is a little foggy as it was 7 years ago), I believe we mostly switched away from it for maintainability reasons. It wasn't under active development, and we were in the middle of a complete MVC rewrite to impose some order on some otherwise organically grown code.

That version of web.py has little relation to the one that Aaron wrote. The current version looks pretty good but I have no idea what it's scalability and modularity story is.

As a slight aside, what do you usually do to stress test a new framework to see if it's scalable?

Ask the community. :)

That's a little glib, but basically I ask other people with experience. You can throw as much fake traffic as you want at a framework but there is nothing like a real world test.