[web] google.com source code

Perhaps they want to keep browsers in a particular quirks mode that works for them?

Google is doing more weird things with their HTML source. It's like they're even trying to keep compatibility with IE3, NS3 and the like using HTML 2. See this.

Keeping support for old browsers is hardly a reason why Google has bad HTML. The most probable reason is that it is not cost-effective to do so. If for example they do change the code live then Google will need to stop the site from running for a couple of minutes. The main problem is if the code is changed and changed incorrectly then there could result in some major problems, which will loose Google money.

One important point to mention is if the code is changed and becomes valid how many new customers will they receive; probably very little if any.

I seriously doubt that Google's architecture requires them to take the site down globally while they make a tiny HTML change. They make changes to various pages all the time and availability is pretty much constant. Their load-balancing system is more than capable of handling such roll-outs. Nor are they likely to have any sort of trouble in migrating to a more modern standard in minimal time should they so wish, given they have half of computing's best minds working for them.

I'll have to agree with Kylotan--even if they did screw up somewhere in their update, they probably have an offline development version of their site. So they'd likely catch any problems before going live.