(It's just not something I think about very often, and only now do I realize how useless my knowledge is! I could speculate based on the flux of high-energy particles and the layout and energies involved in various storage media. I would expect the energies on non-volatile storage to be much higher than volatile storage, and so for bit flips to be rare. I would expect flips to be singular, and not come in groups - but if they did it would be temporal, not spatial - something like a solar flare. The interesting flips would be the least likely ones, where the serialization code is running. The threat-model here is the relatively large memory region consisting of the union of something like [cpu, kernel, jvm, thread, bytecode]. That's quite a surface area, and while it's fun to speculate about what a bit flip could do at various interesting points, but I assume for simplicity they all just crash your code. (I wonder how often random bit flips are why restarting software is unreasonably effective at fixing problems - or is it just an epidemic of thoughtless race conditions as I've always assumed?)
(I. E. hit save, create email , attach the file, does not work. Must close PowerPoint first).
I imagine other usability issues exist which are similarly much more frequent than cosmic rays.