Much of the external circuit is clear. The code 1B for a SOT23 part on the bottom side indicates a simple NPN - so no external JFET stage.
How the chip works internally is of cause not visible. It is even likely that the paths join before the variable gain part, as the variable gain can be quite some effort.
The coarse gain steps are externally, though only about a 1:7 voltage ratio for the 50 ohm path. This does not direct match the gain steps for the user interface.
The 50 ohm input does not need much attenuation, as the maximum amplitude is limited anyway (possibly more than with other scopes).
For the 1 M path the later dividers have no matkings, the first seems to be 1:4 (750 K from the input side).
-snip
The front end chip definitely chooses between the different inputs. It looks like 2 x for 50 Ohm and 4 x for 1 M mode + an additional DC input that also has to be somehow combined with the AC part !
-snip
So the simple answer to my question is : Yes, we do not need a double path to ADC to provide 50 ohms input capability.
A agree with the yes, but not because the signals all go to the same chip.
The main point is that to combine the speed of both ADCs, one needs to send the same signal to both ADCs. So the 2 outputs must be equivalent and to both ADCs.
Similar the same PCB means the 4 unused links have to go to the unused ADC - so the 2 outputs are just to serve 2 ADCs, not to combine the 1M and 50 Ohm part later.
So enabling the 50 Ohm mode should be a SW thing only.
There may be a good reason in no enobeling it with the cheaper version: the 50 ohm inputs are somewhat prone to overload. With this scope this could mean damaging the front end chip and not just burning a 51 ohm resistor.