I then deinterlaced using fieldskit with these settings:
Timing: Motion Estimated Blend
Field Order: (depends - upper for beta)
Fill Method: Blend (other settings gave artifacts)
Detect Motion: 3 Frames Compare (some minor artifacts with 2 frame)
Detect Method: Point Difference (didn't see much actual practical difference with other setting, though mask varies greatly)
Note that the original footage here came though an unknown process before it got to me... so maybe that's why it (possibly) helped. The "unknown process" was something like Film -> ? -> Beta -> FCP over Kona LSe component. I got the Beta tape and saw the interlacing on the ingested footage from there.