What I did was upgrade to vB3, which incorporates some things like thread preview, but it comes highly optimized already. It's got deferred thread views, attachment views, etc. all inbuilt.
When I ported some of my custom hacks over, I've had to modify the code of the hacks substantially to make them better as they pushed the stock vB3 load higher. I've learned never to use anyone else's code without checking it through and optimizing it first.

So far, the loads for my vB3 is high at peak times (around 5) but that's with 500 people online at once.