Gavin was testing his proposed changes, that include increasing the max block size from 1 MB to 20 MB.  He noticed that the nodes  were taking too long (hundreds of milliseconds) to validate a block, even when the node had seen all transactions before and thus did not need to validate them again.  That wasted time impacted the time needed to propagate a block from node to node.  He tweaked and hacked the code until he got that part of the code to run a lot faster, saving some memory too.  That improvement reduced the estimated 1MB-block propagation time, in that easy case, from 600 ms to 10 ms.

That is good news for people running nodes, but it will not be noticed by ordinary clients. The target block rate will still be 1 block every 10 minutes.