It depends on your data, and how you are using it... Worst case, it should perform similar to SQL or slightly worse. If you can reliably shard your data, and don't need result sets larger than 16MB (not sure if they raised this in 2.4), you can do pretty well. Again, it depends on your data, structure and use.

In some cases a traditional RDBMS is a better call. In others, for mongo, having an additional collection with your indexed subset may be the best approach. It really does depend. I don't consider it best for all cases, but do consider MongoDB a really good fit for a lot of web based cases.