Just assume that it's compromised and generate a new one. There is no point in wasting time trying to estimate how long it might take someone to crack it.
It may also explain some suspicious behaviour / source of compromise in the past (we know when the issue was uncovered, not when the first dump was taken)