I'm not sure what dabeeeenster had in mind, but I know that we previously integrated third party forums and other modules into our servlet based applications. You still have the option to run servlet based components on a subdomain to keep your core Play app clean.
I think because Play is Java, you tend to want to merge it into existing servlet applications, but you have to realise that it's the same as moving to an entirely new language. It might be based on the same core language, but its not really compatible.
Although we were excited about creating something with Play, we had to hold back several months because there was no way to gradually migrate and integrate Play into our existing servlets. It only really works when you are starting a new project. Its not ideal, but as previously said, you have to treat it as completely separate, not linked to servlets and its as though you are moving to Ruby etc.
I don't have any systems deployed to production yet, but am in the process of creating an internal app using Play and am liking what I see so far.
There are a few minor problems with the framework being quite immature, and I don't always think the creators have a clear roadmap of where they are going. However in a way this is good for me, because the early adopters can help drive things, but with the core team trying to keep the framework as light as possible.
Coming from a j2ee type background, running each instance as a standalone process, not inside an appserver was a little strange at first. But the process separation has turned out to be a good thing for me, just takes a little more management. But because things are structured in a REST/share nothing architecture, the ability to use a proxy with failover configuration and doing hot deploys via this is nice.
The team are very responsive with help on the google group, and bugs are fixed very quickly, you just have to be prepared to run with nightly builds to keep up to date.
One worry I do have is how the team/community will continue to handle if things do become more popular. I am worried about an influx of new developers clogging the current direct channel to the developers and feel the community will have to split into newbies/users/developers to handle things more efficiently. But they are doing a great job so far so long may it continue.
I think because Play is Java, you tend to want to merge it into existing servlet applications, but you have to realise that it's the same as moving to an entirely new language. It might be based on the same core language, but its not really compatible.
Although we were excited about creating something with Play, we had to hold back several months because there was no way to gradually migrate and integrate Play into our existing servlets. It only really works when you are starting a new project. Its not ideal, but as previously said, you have to treat it as completely separate, not linked to servlets and its as though you are moving to Ruby etc.