I was tempted to recommend MySQL based on the fact that what crazy described is actually quite a good fit for MyISAM's performance characteristics, as far as I understand them. And that I look after a ~100Gb MySQL replicated database.

One thing I've discovered about the replication is that quite happily, it is at SQL level. This means problem solving when it falls apart (which it does do from time to time) it is because of an SQL-level problem which only takes SQL skills to fix. Unfortunately, it is at the SQL level, which means replication basically re-parses and re-executes the update queries. Which also means that SQL errors can stop replication. But on the gripping hand, you can do wicked stuff like having different storage engines as an extreme case of different configurations...

Wade.