I recently managed to buy on eBay and subsequently resurrect a Tektronix 2440 scope (PSU caps). Although the NVRAM chips were still working, they had a datecode from 1993, so I decided to replace them. I ordered two DS1230AB-70 for replacement (those suckers cost nearly as much as the scope). I desoldered the old chips, which was a fortunately a fairly easy job.
Since I installed sockets for the new chips, before I transfer the content from the old NVRAM, I had the opportunity to confirm or bust the myth showing up on many forums that loosing the NVRAM content will screw you, unless you have the expensive equipment to recalibrate the scope.
Actually, there are two NVRAM chips (DS1230AB-120) U350 and U664, both on the CPU board. U350 is used to store the saved waveforms, so if the battery goes flat in that, the only inconvenience is that the waveforms wont be saved when the scope is off. Also, it will enter extended diags at every boot, as the waveform data is invalid.
U664 is however more important. It stores the calibration data and settings. The service manual is not entirely clear on what happens when this data is lost, but it is implied that a self-cal and an ext-cal will bring the scope back to spec after a cold-start. All the other scary adjustments (many involving turning pots) in the service manual are only required when replacing stuff like the CCD modules. So I figured that the pots are there where they need to be, I'll be able to recalibrate the scope.
I put in the new chips and fired up the scope (not forgetting to remove the calibration enable jumper first). It indeed stopped at the self test. Surprisingly it was smart enough to figure that the NVRAM is empty, it did not even try to run the above 6000 tests, as they would likely fail without the calibration constants. Instead it told me to run self-cal and ext-cal once the scope warmed up.
While waiting for the warm-up, I did experiment with it. The scope was badly out of spec, most evidently the attenuator was all over the place, there were gross errors, on one of the pictures you can see where the scope thinks the GND is and where it is really.
After warm-up, self-cal was run, as the service manual warned, it did take much longer with no initial calibration constants. Then I set up for ext-cal. The attenuator and trig cal is easy, you just need to apply specific voltages to the ch and trig inputs, the scope walks trough you on this.
The channel delay and the CTE cal is more tricky. The service manual doesn't really specify what signal the scope expects, other than it is from the fast rise output of a specific calibrator. This could probably be figured out from the manual of the calibrator.
So I used my pulse generator. The delay time calibration is pretty simple, the scope was satisfied with a 100kHz square wave. For the CTE calibration it is more picky on the amplitude and rise time, but after some fails, I eventually managed to pass.
With warm-up time included I spend under an hour on all this. I know, because the cold-start resets the hour counter in the scope.
Obviously I do not know if the scope is really in spec, for that I'd need equipment which I don't have, but some initial test don't indicate anything out of order, for my hobby purposes it's good.
So the TL;DR of all this is it's not the end of the world if you loose the NVRAM content in a 2440, with some fairly simple equipment you can restore it to working state.
Attached are some (sorry bad quality) photos on the internals of the scope and some other pictures relevant to this post.