In reality, not much. I imagine this sort of thing has been going on for years, but only recently have I been around people knowledgeable enough to call them on it.
Although I disapprove of that sort of price discrimination when it is taken advantage of, which is easy enough for a manufacturer to do, there is another way to think about it.
The additional costs of manufacturing several different versions may make them less economical than manufacturing a uniform design. By offering price reduced models with identical hardware, the manufacturer provides a lower performance instrument to those who would not otherwise be able to afford it. This applies even more now to programmed capabilities like decoding which have considerable development costs; when you buy the option, this is what you are paying for and it might be considered unfair to charge everybody for it even at a lower cost to everybody when they do not actually desire it.
Do not discount the testing, calibration, and grading issues.
Testing and calibration can take significant time adding to the cost. For an example of this consider the LMC6081 operational amplifier and the identical LMC6001 operational amplifier. The only difference between them is that the LMC6001 is tested for a much lower maximium input bias current which takes an extra 10 seconds resulting in a part which costs several time more only because time testing is expensive. Many of the old Tektronix TDS series of oscilloscopes had 500MHz and 1GHz models and the only difference between them was calibration, which even while automatic, took considerable time. If you hack the Tektronix TDS 500MHz models to 1GHz, they show poor (useless) transient response unless the time consuming 1GHz calibration procedure is done.
Tektronix used to grade 2N3904 transistors for collector-base time constant (rb'Cc) for use in their 100MHz oscilloscopes which is why I found it interesting that Rigol used 2N3904s in their vertical amplifiers which Tektronix never did; they used their fast 2N3904s in other areas of their 100MHz instruments. Is Rigol grading the transistors or finished instruments for 100MHz performance? It might explain the non-linearity I pointed out above in the bandwidth hacked instruments but I have not seen the same test on done an official 100MHz DS1104Z.
Below is a photograph from an earlier discussion where I marked the problem and you can see the same thing in the previously linked discussions. That unusually linear feature in the settling time (1) should not be there; it may not look like much, but it even exceeds the transition time of the oscilloscope! That is *not* 100MHz performance no matter what the measured transition time and bandwidth are.
Take another look at the reverse engineered schematics of the DS1000Z series that Dave made. That switched differential amplifier concept was used in many analog oscilloscopes as an additional switched gain stage but not here; the same emitter and collector loads are used for both differential amplifiers producing the same gain. But the emitter networks for equalization are different. (2) Take a closer look and there is something even more odd; the tail current of each switched stage is adjustable! That is not a simple selection between two different equalization configurations; it is more like some kind of calibrated peaking adjustment which is exactly what I would expect if the 50, 70, and 100 MHz models have separate calibration requirements. (3) And if the differential tail current was set low, it could result in the waveform problem mentioned above. Low tail current is actually something I would specifically look for if I saw that.
(1) If you refurbish and calibrate enough oscilloscopes, you start looking for things like that which indicate a real albeit obscure problem. Another problem would be bandwidth which varies with vertical sensitivity.
(2) Unfortunately Dave did not record the capacitances.
(3) Analog oscilloscopes make this adjustment with capacitive and resistive trimmers where those series RC emitter networks are (4) but if you want automated calibration, then you have to do something else like what apparently Rigol did. Adjusting the tail current does affect the -3dB bandwidth but it affects the transient response even more and I have never seen it used to limit bandwidth and I do not think that is what is going on here; some Tektronix oscilloscopes adjusted tail current for maximum bandwidth of large signal stages which is a completely different situation.
(4) The Tektronix 2235 series of oscilloscopes came in 100MHz and 60MHz models. One of the major differences between them was the existence of those equalization networks; they are completely missing on the 60MHz models although the boards have the spots for them. You can remove the bandwidth limiting network but the transient response will never be correct.