Scheduler converts coordinate to JNow before sending it. As I said before everything in Ekos is in JNow. The astrometry.net results are in J2000 but they're converted to JNow. If you see any discrepancy in the logs regarding this, please report it.
Tonight I had a weird issue that may relate. Because of another problem, I had much guiding deviation while guiding on M52. Scheduler finally gave up, and moved to NGC281. Strangely enough, it kept trying to align to Sigma Cas instead of the expected target. Plate solving did its job, but with wrong coordinates. Calibration was cleared, same result. It appeared the RA/DEC that were supposed to match NGC281 did update for whatever reason in the Scheduler interface, and were now wrong. I even reloaded the original .esl and coordinates were still changed to Sigma Cas when starting to plate solve. I don't know what to think, it was quite late, but it seems the RA/DEC coordinates of the SchedulerJob object can be updated by something else than the Scheduler.