What FR really needs is a clustered database, that automatically replicates data across servers. Alternatively, use of a distributed cache like memcache would probably make a big difference. It's how large sites like Facebook, Wikipedia, YouTube, Reddit, and other similar websites handle large amounts of dynamic data.
There's no simple solution. Faster servers can help, but you run into other limiting factors if the application can't handle high levels of concurrency.