This was the first any only time the bitcoin core has been hacked (please tell me if I'm wrong about that). And it was spectacular. At 6:08 pm that day, Jeff Garzik (on the bitcoin core dev team and CEO of bloq.com) reported that "The "value out" in this block #74638 is quite strange: 92,233,720,368.54 BTC?" An attacker had taken advantage of an unforeseen check transaction overflow error.
2.5 hours later Gavin Andresen had a patch to fix it. 30 minutes after that Satoshi posted the patch the devs ended up using. But how do they implement this major change in the code (which includes instructions to ignore all existing blocks after #74637) without hard forking to a shorter version of the block chain? Satoshi asked everyone to stop generating blocks, but it was pointed out that the community was just too big and distributed by this time to expect quick voluntary action on anything. Satoshi then gave step by step instructions on how to create a new blockchain by applying the patch and then downloading a backup of the blockchain to block #74,000 and restarting the client to restore the remainder of the blocks, ignoring anything after the last "good" block.
Then it was a race to generate blocks faster with the patched code than the original block chain, which was still generating blocks of its own. The core code with the longest chain would ultimately win. This was no small task. Implementing the new chain with the patched code was not as simple as pushing an "Update" button. It would take some time and effort from a lot of miners to fix this problem.
By 2:00 am the next morning it finally looked as though the new block chain had more hashing power than the old one. After just 50 blocks were generated, the new chain took the lead. Fortunately there were devs and users active on the forum who had a relatively large amount of hashing power.
I find it fascinating how brittle the early days of bitcoin were. I had just always assumed that once the core was published, it was a Satoshi Magic done deal.