What is the worst case slope on the periodic error curve that ZWO included with your mount?
From the first posting in this thread, Dave has one of the better ZWO mounts out there, with magnitude of slopes similar to my RST-135. My RST-135 has larger p-p periodic error, but smaller harmonic distortion. That is why I jumped into this discussion with more "color."
As in everything in this hobby, don't blindly use some other person's number. Measure that slope carefully and then set up the max pulse duration. If you have one of the worse ZWO mounts, as I mentioned earlier, there is nothing you can do.
For example, if the slope is over 1 arcsec/sec and you are using 1 second exposures, then the RMS guding error (just of the RA axis) could never be always less than .35 arcsec. If the slope is worse than that, the RMS error will also be worse.
The only thing you can do is to reduce the exposure time to 0.5 sec by using a reasonably sized guide scope. And if the slope is still too large to match your plate scale, then you will just need to move to a different mount.
By the way, also check where your error is coming from, the RA axis or the Declination axis. There were some postings in this forum complaining about the large backlash in their ZWO mounts (the strain wave gears do not have much backlash, but the rest of the mechanical design can add backlash).
If you are driving the guide rate at 0.5x sidereal rate, you can never get a backlash in the RA axis (except when you dither)since the RA axis is being tracked at 1.0x sidereal. A 0.5x sidereal pulse will only change the track from 0.5x sidereal to 1.5 x sidereal, and since both are positive numbers, the pulses can never produce a backlash.
The story is a little different in the declination axis. If you are seeing a large declination guiding error, introduce some polar alignment error (that's right, you read that correctly) so that there is a declination drift rate that is larger then the slope of your declination axis. Unfortunately, ZWO does not publish curves for them, so you will have to use some trial an error. If you remember, a polar alignment error can be measured by watching the drift of a star in declination (check out "drift alignment" on the web). So, just add enough polar alignment error so that the declination drift is greater than the curve of your declination gears. Your exposure duration may be limited by the field rotation, but that is what one has to sacrifice when there is lots of error in the mount.