Spent some time comparing my 53230A counter to my 33521B last night. The PLL/clock systems aren't that similar between the two but I was able to find enough in common to make an educated guess on how we might get the AWG's to recognize they have an OCXO installed and to also operate properly. After some experimentation I believe I have figured it out.
It is nice and simple - remove the stock osc, install the OCXO and then remove resistor network R901. Removing R901 enables the OCXO in firmware and hardware at the same time.
You can then power it up and wait for things to settle down to perform a reference calibration. The calibration is easy and are the first steps in the calibration routine, you can exit when the reference calibration is done so as not to mess up any other settings.
The OCXO I used is a Vectron C4550A1-0213 - it was a whopping $13.99 shipped for a surplus one out of China.
Keep in mind this mod is only as good as the OCXO that is used. The original Keysight one is a Vectron C4550A2-xxxx I believe. The "A2" is a taller package with more insulation vs the "A1" I used. The last 4 digits on the end detail the specific options ordered but most aren't covered in the datasheet. Vectron makes all kinds of variations of this model, AT and SC cut, different stabilities etc. I am quite sure Keysight uses a very good SC cut model. So just because we add the OCXO doesn't mean it will meet the same specifications listed in the Keysight datasheet.
So far mine is working well. When I connect an external reference I do see a brief "PLLERR" before it locks to the external reference. I am quite sure this is normal. This isn't an error that is logged in the official error queue which is why I think it is normal warning, and it is quite common for things to take a few seconds to synchronize to an external clock. We can only verify everything if we eventually find a factory OCXO model to compare to.
Here are a few pictures to assist: