So after 4 lost perfect nights spent troubleshooting this issue, the root cause ended up being PHD2.
I don't have a specific answer as to why it was causing this problem but switching to the internal guider ended up resolving this issue.
Full troubleshooting journey at the following:
Everything Is Not Always Great!

and
The Most Bazaar Astrophotography Problem I've Had!!

Thank you Douglas for taking the time to explain your issues. I'll check if I can reproduce this issue with PHD2 though I have to say I use the internal guide exclusively.

Thank you Jasem!
The main reason that I used PHD2 was that it was running on the RPi where my instance of KStars/Ekos runs on my Mac. I was concerned about network latency and guide pulses. This, however, does not seem to be problematic at all!
The other consideration has been predictive PEC. Does the internal guider provide similar PEC functionality?
Cheers!
Doug

Yes, it does. It's called GPG. Click Options (near the right side bottom) in the Guide TAB.

Thank you Alfred! I do have GPG enabled.
Do you know if there is reporting/graphs to analyze periodic correction?

The core of the GPG code is from Edgar Klenske's PhD work, e.g. ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7105398 and (at least as of 2 years ago) is the same code as he contributed to PHD2. Of course there are wrappers etc to include it into KStars which may be different.

You can export an Ekos guider log, which is output in PHD2 guider format (on Linux/RPi look in ~/.local/share/kstars/guidelogs) and you can use the PHD2 log viewer to analyze it. The guider log is enabled by a checkbox all the way at the bottom of the Guider Options menu (Guide sub-tab)

Hy

Thank you Hy!
Oh that is interesting and good to know!
It looks like internal guider user log was enabled already so I should be collecting data.
Cheers!