Thanks, that gives me some hope -- One less data point for the mysterious corruption hypothesis. (Although my problematic keystore file is one byte smaller than the new ones I generated today.)
To be honest I can't be certain what password I used, although I'm sure I chose something "obvious"
(All of the "keep this key really safe" stuff in the docs dissuaded me from writing it down.) Unfortunately the script I wrote to try all my usual password permutations hasn't found it.
I wish the error message was more definitive -- It has real red-herring potential