Well, bit flips happen all the time, and we call that "cosmic rays". It is not obvious that bit flips from cosmic rays in the more usual sense of "cosmic rays" happen all the time.
That's actually testable and determinable from several perspetives.
- Isolate any other sources of ionising radiation.
- Check to see if prevalence increases or decreases with increased or decreased cosmic ray exposure (atitude, shielding, detected cosmic ray storms).
- Are the characteristics those of single-bit flips? (E.g., power-of-two changes to values, or similar.)
- Are the errors nott repeated for the same hardware component. (E.g., under increased cosmic-ray influence, bits flip at an increased rate, as predicted, but which bits flip is random and has no detectable pattern, as predicted.)
You end up with very strong circumstantial evidence of altered bits due to cosmic ray influence.