About at height 10400, rigel started his script to pump out transactions. If you are at 10081, the next batch of blocks you request from a node goes up to about 10700. There are many transactions to fetch from the db. That's why you run into the timeout if you are unlucky. We have to look into that issue. There seems to be a bottleneck.
So his script caused this? Perhaps once this instance is fixed we should get a number of people running scripts to test it even harder? Can't let this happen after launch. We should try to test the network beyond the amount of transactions that are expected after launch...
It uncovered a few bottlenecks in the code that we're working to fix.
Any ideas yet?
Sorry, but Im really curious about that ;)