This has happened ever since I purchased the light (in 2004) and after I got a replacement TC by warranty (in 2006) because the lock-out feature no longer worked (if anything this is indicative of how often I used the lock-out feature
). This happens irregardless of brand or freshness of the battery (including SF), and the only other brand I use is PentagonLight (which made great batteries).
This isn't the first light that had this happen to me. I have a Pila GL3 (old style) and a PL eL2 that ends up with dead batts if they sit too long.
I guess it's just better to leave the batteries out under long-term storage.
I don't suppose you'd be interested in doing an experiment? Specifically, take two sets of cells from the same batch (same brand, same manufacture date, fresh from package, etc.). Put one set in the A2, activate it on high for a given period of time (maybe ten minutes?), then take out those batteries, and set them aside, leaving them untouched for six months. Then, with the second set, put them in the A2, run it for exactly the same amount of time on high as with the first set of cells, but this time keep them in the flashlight, and set the light aside for the six months. At the end of the six months, run both sets of batteries dry on high on the A2, measuring how long each lasts. Assuming they're good cells, then this should tell you whether the power drain is a CR123 issue, or something specific to your A2.
Of course, such a test would put your A2 out of commission for six months, which may not be possible depending on your lighting needs. But if your use of the A2 is pretty infrequent, then even just swapping out the "test" batteries for "general use" batteries on the occasions when you need the light, the test should still be useful, so long as the "test" batteries go back in as soon as you're done using the A2.
Granted, this wouldn't be the most reliable test setup, but if it yields a significant difference between the batteries left in the flashlight vs. the ones stored separately, then you know it's something to do with the light. And on the other hand, if the cells perform the same (and exhibit the same reduced capacity you saw before), then maybe the issue of CR123 self-discharge after partial depletion deserves further scrutiny?