I question your assumption that something from the switching supply is somehow getting back through the audio connection to influence the VFO control voltage. You could simply be hearing IM distortion from the audio amp.
Hm, that's a good point. Come to think of it, the standalone amp unit doesn't have any voltage regulation. So in addition to IMD the audio amp supply voltage instability may be causing the warble.
Regens are known to be hypersensitive to supply noise; you should probably banish any switching supply from your bench while developing... and I don't know why you'd even consider one for this sort of radio project.
Well, I have one around because I bought it to power my yet-unbuilt QRP CW transmitter kit (VXO), as the manual specifically showed a switching power supply being used in the construction photos. I was hoping I could use it for my receivers as well but am starting to understand the difficulties.
Re your requirement for stability - what's the final goal?
1) To determine if I need a superhet for on-air CW work, given my location and antenna. My finished receivers until now have been regens. One thing I read a while ago here on TheRadioBoard has been gnawing at the back of my mind though: the fact that a good DC or superhet receiver can hear signals inaudible on a regen due to the regen's higher noise. That bugs me, that a better receiver might hear more than my regens. So one goal is to build a superhet with reasonably modern specs for IMD, sensitivity, noise, etc., then to see if at my location and with my antenna I can actually hear any more signals with a good superhet than with my regens.
2) If the superhet is actually better, then to use it for on-air CW work. This mandates crystal-filter selectivity (both to reduce audible interference as well as to protect the sensitive IF strip) and sufficient VFO stability to prevent signals drifting out of the passband.
A third goal is, having gone through all the trouble of building a good superhet, I'd want to use the set for general coverage SW listening.
Your suggestion about a frequency counter is a good one for simplicity: Just live with VFO drift, and use the brain-locked-loop to bring it back on frequency.