ET and solar here. I had something similar going on. My solution was suggested by a nincompoop pool guy, and somewhat rejected by a solar installer, who hadn't heard of it, so logic should dictate that it was a questionable idea. But it did work.
The pool guy explained that when the temp reading on the roof is very close to the "hot enough - not hot enough" threshold, any inconsistent wind on the sensor would affecting the reading, first pushing it a little above the threshold, then cooling it a little below. This would cause the ET to react, first engaging the solar, then pausing it, back and forth as the gusts affect the sensor reading. He suggested a glass container mounted over the sensor would not affect the temperature reading caused by the heat from the sun (or the roof), but would negate the fluctuations caused by the wind.
I've been running it like that for years now, the on-off-on-off problem vanished, and the system seems to be working as intended. This may or may not be a solution for you, but it's easy and cheap, so might be worth a try. Just remember, this is not a standard solar installation technique, but rather an anecdotal trick, made up by who-knows-who.
I used a small pickle relish jar, including the lid. I hacksawed a small slit in the lip of the metal lid, just wide and long enough to allow the sensor's wire to pass through into the jar. Then I mounted the lid to the roof, using a screw and some roofing sealant, inserted the sensor into the jar, routing the wire through the slot in the lid, and then just screwed the jar, upside down, into the lid. It's all pretty simple. I considered using a plastic container, but then decided to use glass because it would withstand wind and rain and UV, and minimize any interference between the sun's rays and the sensor. I think I wrapped some layers of tape on the wire, where it would have contacted the hacksawed metal, just to give the wire a little extra protection against any abrasion from the sharp metal edge.
The logic behind this makes sense, blocking the wind from the sensor. But I never experimented to determine if the temperature sensing is affected by the jar. Is it hotter in the jar? Or cooler? And if either, does that matter? All I know is that the system works fine, and the on-off-on-off issue stopped happening.