Well, I can invent scenarios that could explain the observations; whether they're correct is a different matter altogether.

For example, suppose these DRAMs aren't truly hermetic. Over time, moisture could seep in and corrode things (say, a bond wire or bondpad). Or if gold bondwires were used, it's possible for the dreaded "purple plague" to work its nefarious magic.

These problems would show up over time in powered or unpowered units but if poor hermiticity were the problem, storage in a damp environment would certainly be a bad thing. An operating machine would also be a hot machine, perhaps keeping moisture somewhat under control.

Without some failure analysis on the dead chips, one can only speculate. But mechanisms that could plausibly explain a couple of dead DRAMs in one machine do exist.