saguaro sure. I find that at the beginning of a stargazing session, if you have not specified a target via the sky Atlas, and it starts performing the horizon calibration by that 3 points sky calibration and successfully finished that step, it will try to slew the original object. But since no original object was ever specified, I find that it slew to what seems like a random spot, even though in theory the 3 point calibration should have known the exact location of that center spot. So even if the section of the sky is clear so the calibration works, of that the view to that random spot is somehow obstructed, it will fail and thus default to the older leveling algorithm.
This was repeatable in my end as I was testing in my backyard where the back of the house blocked much of my sky. That is why I suspect ZWO tested this algorithm in an open sky area so these "edge cases" we're not examined.
Hope this helps.
Kai