This sounds like a bug.
There's a chance it could be JBESB-1631, but I doubt it. It sounds more like the redeploy is stopping jobs that it shouldn't be stopping. This might be because the jobs are not being uniquely named.
First off... are you in a position to test against a build of the latest CP branch of the 4.2.1 codebase (CP2). You'll have to build this from source, but it might be worth doing, just in case the Scheduling related fixes there fix your issues.
After that, you could get the 4.2.GA code from the SVN tag for that release and step into it with a debugger. I'd look at how the jobs are being registered and unregistered (in the ScheduleProvider class I think) and make sure they're being registered with unique names and unregistered appropriately.
I should also add that moving to 4.2.1 or, if you can wait a few weeks, for the next release would be a better alternative.
Is there a reason why you are still using 4.2GA?