production servers should be stable (unless there's a huge available pool of personpower to do q/a testing with specific snapshots)

but for a laptop, I think it makes sense to run unstable (or "sid" in debian-speak) for people who are interested in learning -- for a couple years I was able to track unstable on about a twice-weekly rate with my laptop with only very infrequent breakage -- right now I only do "aptitude upgrade" every couple of months or so (and it kills most of a day getting all the available updates and reconfiguring them -- so far the 2 times I've done this with my new laptop haven't been problematic at all)

which brings us to the next strong suggestion -- get apt-listbugs installed -- it's really helpful when tracking unstable, as it alerts you to open bugs in packages so you can put off installing them

I've never been a big fan of testing as an environment to try to keep a machine configured with -- certainly during periods where one is looking ahead to the next release, and before there has been a freeze, it makes sense to keep a sharp eye on what's happening in testing (especially as far as it relates to whatever applications you'll be installing on the server you're thinking about bringing up in the future).  On the other hand, there are people I know (some of them much smarter than I) that do advocate using testing as a base for production servers.