Switchers can certainly be twitchy beasts, can't they? Congratulations on figuring out the Zenith problem -- that's some mighty fine troubleshootin', my friend.

I once came across a problem that was the inverse of yours -- a prototype supply worked great with cheap caps, and the company went into production with more expensive ones without doing any electrical tests (it was "obvious" that it would work even better). Turned out that the higher ESR of the cheap caps was needed to stabilize a poorly-designed feedback loop. The better caps, with much lower ESR, caused the supplies to oscillate into oblivion. Took a lot of head-scratching to figure that one out.