Speaking from direct experience, faced with the job of developing a framework for web data entry apps, what I ended up doing was writing a dataflow language for data binding. Another ex-Delphi compiler developer, Danny Thorpe, ended up co-creating Google Gears. IMO compiler development is a great basis for a certain kind of thinking about solving problems; but I'm also biased.
EDIT: Sorry, did some research and assuming its this: http://en.wikipedia.org/wiki/Programming_language_theory
Actually, I was at a presentation today for new grad-students, and one of the older grad-student mentors helping to present mentioned that the coursework phase of grad-school is a good time to get a bit of industrial experience in the summers. Once I actually have real grad-level PL experience under my belt I'd quite like to go and see what it's all about.
If you hire lots of, I don't know, compiler specialists but you allocate them to write web apps we're all possibly worse off. The issue is, and what you're not cottoning on to, is that in a company the size of Google there's probably a team working on x for all values of x that could maybe use another member.
They're only tolerant of their resource allocation scheme because they probably don't worry about their employee acquisition cost.
With a company the size of Google, it'd be really interesting to deregulate their internal labour market (i.e. what Valve does). Let people switch all the time! If you need to prioritize one boring, yet necessary team allocate higher salaries to it.