If the GPSDO has a weakness at the moment, it's that every once in a while, GPS decides to go whacky for a few minutes. I see the frequency difference trace suddenly go 3-5 ppm in one direction, then come back 6-10 ppm the other direction, then 3-5 ppm back to where it started - all of this over the course of just a few minutes. Often this goes hand in hand with a marked rise in PDOP for the duration.

In principle, it's possible to add a firmware workaround... If you have the frequency dialed in and it suddenly makes a large excursion, we could give the GPS receiver a "time out" for 5 minutes and free-run, with the hope being that when we let it out of the corner it's going to more or less be in the same place.

I'm loathe to do something like that, however. If you don't get that hack exactly right, you could wind up making things worse rather than better.