Note that I trimmed the area when I captured an image but neglected to do reflect that in the scaling options. I don't know if that had a deleterious effect.
Also, I ran solve-field on the remote computer, blind, without any options at all. Although I didn't record how long it took to solve, it was quite rapid. So I wonder if running the solver remotely created some kind of bottleneck.
There also seemed to be a problem with solving to coordinates for which Ekos had no object. Thus, for example, I tried solving to a star dimmer than 8.4, but the solver solved to a near by brighter star. But I'll investigate further and post separately if it continues to be a problem.