1. If you delete data, tombstones are a problem that nobody wants to have to think about.

2. Denormalizing data instead of joining isn't always practical or feasible.

3. If you're not dealing with large amounts of data, Postgres and MySQL are just easier. There's less headache and it's easier to build against them.

4. One of the big advertised strengths of Cassandra is handling high write volume. Many applications have relatively few writes and large amounts of reads.