I went to school that primarily taught COBOL/MVS assembler (with hints of C/C++). While I didn't enjoy the development environment as the tools seem so rudimentary in comparison to more OO based languages, the mainframe is still an impressive technical beast. The amount of raw I/O mainframes can shuffle around and process is unbelieveable. x86 servers still cannot replicate the mainframes raw power (nor cost).
I did do some professional work on the mainframe. Most often I was reverse engineering (CICS/COBOL/MVS Assm.) stuff so I can't say it was development, but nonetheless it was a good experience. I have a feeling that people tend to go for what they find is easiest and if it just so happens to be interesting to them, all the better.
In terms of your questions:
Will this be a hindrance when I try to
find a new job (web development, game
development, etc.) or will this
actually look good to a recruiter?
Not too much. If you try to jump fields within programming you may have to start at the bottom, but it won't hinder your chances of landing a position. If anything learning the mainframe will probably make you more marketable within industries that still use it such as government, banking, insurance, etc. etc. So really, programming (in any language/environment) is largely about the industry you're working in. The industry will determine the technology/tools. If you understand the core concepts of programming, jumping industries will probably require you to learn newer tools and more concepts, but I don't think anyone would hold it against you for a position. Granted, if you switch, it'll be like you're just getting out of college again and start at the bottom.
Would I be locked in the financial
sector and big companies, or are there
other sectors who use Mainframes (I
only hear about the financial sector
all the time)
For the most part, yes. If you work in the mainframe now, it's likely you'll be locked in later. I've met people who've been working on the mainframe for over 25+ years. Don't tell me that's not being locked in. It may be a golden cage/cell.. but it's a cell nonetheless.
I've worked in the insurance industry. Insurance still lives (and dies) by the mainframe. For the most part I can't see them fully leaving the mainframe but the associated costs are very high and I know (from experience) that some insurance companies are trying to.. alter their usage of mainframes. If you know the mainframe (COBOL/CICS) and say.. Java/Web Services type stuff. You'll be the bridge that they'll look to cross someday. And that's a payday worth working for if you're up to it.
Is it a useful investment of my time?
This is subjective. 'Useful'? Yeah. Probably. 'Investment'? Definitely. But if it's something you don't think (or know) that you'll enjoy, it's torturous. If I had to stay in the insurance industry and work on the mainframe, I don't think I'd mind it, but I would pursue other languages and technologies in my own time. You can't stand still in IT whether it's programming or systems administration.
Bear in mind there are risks when placing all your eggs in one basket. Granted my scenario is probably impossible but maybe someday, in the near future, mainframes will be replaced with.. I don't know.. supercomputers or something totally new that puts them out of business. Well.. now you're on ground zero and have to look for something else. That's why if you can learn the mainframe AND java and other supplementary technologies with your mainframe skills, there's no telling where you can go in the next 10, 20, 30+ years.
Just remember, do you what you love and you'll enjoy your work no matter how hard. People who don't do what they enjoy end up leaving and starting all over again somewhere else. Hope my rant is helpful to you (or anyone).