My two cents on the accuracy/stability of the reference in the 2042X. A few months ago, I checked my unit against my HP GPSDO. I checked the frequency error at quite a few points from the audio range, up to 120 MHz. I don't remember the exact numbers, but the error was a linear percentage of frequency across the entire range (which is what you would expect, if things were designed/working correctly). I do remember that at 120 MHz, the error was only a few Hz. After initial warmup, the FG output was very stable.
Yes, this is only a sample of one. However, it does give me confidence that the internal reference in my unit is a very good one, for a piece of mid-range equipment. With this level of accuracy and stablity, I don't see myself hardly ever needing to use an external reference.