I particularly have enjoyed the ORM, although you need to be careful not to get carried away with it. Its easy to create slow queries with it and don't assume it will always get indexes correct.
I have also enjoyed the community around Django. I enjoy being able to search a particular problem and find that someone already has created a Django app to solve it. In my PHP days (RIP) this was possible, but usually ended up being a lot more ugly.
With that being said, I think while development without a framework "might" feel more lightweight, you will pay for it later when you're maintaining your app. As they always say, you build it once but you have to maintain it forever. I think investing in a framework will most likely pay off in the future even if there is an overhead usually associated with framework.
I'm building a digital signage application. I'm using web2py on Apache for my serverside control panel/mobile management website. I'm also using web2py on the display side with it's built in webserver.
To tie everything together I'm primarily pushing around JSON via Beanstalk. All other management of the client ends is done via Fabric.
We tend to think that whatever feature we need, we just google django-feature-x and we find a decent implementation. And the community is great.
we like supertagging, django activity-stream, haystack, django-cities, south, sentry, easy-thumbnails, taggit.
We're about to release a few django apps we've made too.