I've been using Cassandra instance without reboot for a few days for a simple task for storing tweets, 1-2 saves in a second. After then Cassandra got really slow and I had to kill a restart it. Is this Cassandras's expected stability now? Would it be a good solution to write a daemon to kill/restart it every day or two?
No. Cassandra is widely expected to be more stable than that. If it is not stable, there is a substantial chance you have configured it wrong. It may be attempting to use more memory than you expect, for instance. If you have encountered a bug or defect in Cassandra, it is not one which is afflicting the majority of users.
As for your "restart daemon" plan, I'm going to go with "that's a horrible solution for pretty much everything, and especially so for something that you're trusting with any data you actually care about."