So on DS1074Z using 2 channels (500MS/s) we get 500MS/s / 2,5S = 200M/s (200MHz) theoretical real-time bandwidth. Even on 4 channels we still get 100MHz theoretical real-time bandwidth. Even considering that Rigol scopes needs 2 times more samples to create decent result we wont go lower than 50MHz theoretical real-time bandwidth.
So there is no point in using worse method which is ETS and this is why its not pointed in DS1000Z data sheet.
So "25 GSa/s ETS" isn't something that makes DS1052E better than DS1074Z which also can be hacked.
There is a problem with your logic. The 4-channels-ON BW of the DS1074Z is nothing close to a usable 100MHz. Sin(x)/x interpolation is worthless because the frequencies above the Nyquist rate aren't attenuated nearly enough - which means you should really always use linear interpolation with 3 or 4 channels on, making the
functionable real-time bandwidth more like 25MHz.
This 2.5x ratio is considered JUST adequate when you have very good BW frequency response. As a LeCroy document states: "SinX interpolation works very well only when this ratio is greater than 2:1 - 3:1 is a good ratio with 4:1 usually working almost perfectly."
But even then, as this Yokogawa document states, "This formula only applies to sinusoidal (single frequency) signals, which allows the instrument's sinusoidal interpolation feature to be used. For more complex signals, the sampling rate should be well over 2.5 times the highest frequency component of interest, in order to faithfully reproduce the waveshape."
In other words, you have to build a DSO to work well with linear interpolation. That's why a sampling rate of 10x the BW is the standard formula for most modern DSOs. For example, the Rigol DS2000 has a sampling rate of 2GSa/s - allowing a 200MHz maximum BW (and making their 300MHz DS2000A slightly suspect).