Here's an update on what I've found. I took unpotted boards that have been power cycled hundreds, if not thousands, of times successfully and tried v arious tests. (I should note that unpotted boards have never failed when th e input voltage is kept within spec.)
The failures always occur immediately after power is applied, so it doesn't appear to be a heating issue. Here's what I did:
-put RTV (the stuff with acetic acid) around the inductor, and it failed wi thin a few power cycles. I don't believe the stuff was even close to fully cured.
-blobbed thickened potting epoxy around the inductor on 2 boards, but I got impatient and applied power before it cured. Power was left applied while it was curing. Amazingly, these two have not failed. A few days later I potted the entire board in the enclosure and they've been reliable.
-blobbed thickened potting epoxy around the inductor on 2 more boards, but this time I allowed the epoxy to fully cure. Both failed within a few powe r cycles.
-blobbed "electronics grade" (and no acetic acid) RTV around the inductor, let it fully cure, and failed within a few power cycles.
So, WTF? My only guess with the two that didn't fail is since they were po wered during curing, the inductor was allowed to magnetostrict and create a tiny cavity inside the epoxy. But, RTV causing it to fail? I don't under stand.