At DjangoCon, Mark Ramm warned against Django's "Z shaped learning curve". And, he's exactly right. Getting Django installed, pointed at your database, and opening up the admin is a joy. After, that, you have to roll up your sleeves and start learning.
It really takes a while to get used to the work-flow: Design a URL, created your model, created a view to push the data through to the template, design the template. Then, there's the Generic views, which are a whole other language to learn. After you get up to speed, it makes development a whole lot easier, but from the bottom of the "Z" to the top of the "Z" it takes a lot of struggling.
On the contrary with other pythonic frmeworks, the curve is not that difficult. Currently we are using the combo of cherrypy and dejavu in python, and the learning curve was surprisingly small.
We're using sqlalchemy here and it's been nearly painless so far. It scores on maturity (despite the 0.5 version number), great documentation and a very active and friendly community. The learning curve is steep (as with any ORM) but very rewarding because the pieces constantly fall into place. Everyday tasks are idiomatic and syntactically sane, plus even complex joins and subqueries can be constructed using the standard building blocks. So far we have not run into any limitations, unlike with the django ORM which gave us constant headache at some point. The latter really feels like a toy when you have tasted the real thing.
It really takes a while to get used to the work-flow: Design a URL, created your model, created a view to push the data through to the template, design the template. Then, there's the Generic views, which are a whole other language to learn. After you get up to speed, it makes development a whole lot easier, but from the bottom of the "Z" to the top of the "Z" it takes a lot of struggling.