Petar,
Really interested in your ideas. I am concerned that you seem to be looking at the something that is head-to-head with the mainstream products, and your bill-of-materials is very high. Even though you may be able to offer a bit more bandwidth, it seems you will be head to head with the new Rigol 2000 series with its great display and impressive features. And they did promise an "incredible" pricing for the range.
I was really wondering whether there are some more niche products that you may be better off with at the start, particularly if you can start with a cheap BOM.
1. Isolated inputs. This is probably not easy since people will want to use them with switching power supplies. If you have a FET switching 400V in 20nS, that is something like 20 billion volts per second slew rate. If the isolated circuit is powered by a transformer with a 30pF interwinding capacitance, you can still get a 0.6A transient current through this capacitance.
2. Modern Version of the BWD powerscopes. 4 x 600V AC differential inputs. Bandwidth 20MHz or more. Even without much bandwidth, a scope with two separate differential input leads per channel is just incredibly useful. Most of the time, it is much better then the high speed differential probes that are difficult to connect to circuitry.
3. Following on from the idea above, an differential probe like the Fluke DP120. Again, fantastically useful, if only they didn't cost about the same as two Rigol 100MHz scopes. If you could make a $99 probe, you would have a great product. If you could make a $49 probe, it would be a killer product. There is the potential for a very low BOM, so it could be a great way to start.
4. A 12 bit to 16 bit sampling scope. Bandwidth maybe 1GHz or more. 50 ohm 5V input. Sampling scopes are never a replacement for real time scopes, but for the occasions when people do need the speed, they often can be made to do the job. They are great for single frequency repetitive waveforms, or looking at timing windows on high speed digital waveforms driven by a common clock. Also there is the potential for the use of very cheap parts, and the A/D can run at a fairly slow speed, so you can use whatever resolution your circuit can handle. It may be possible to sample with a circuit based on two or four cheap 350pS 0.3pF Schottky diodes. you may have to match your own pairs of diodes. I read recently from a sampling scope designer that you can get up to 10GHz bandwidth using standard off-the-shelf parts (if you know what you are doing). At high frequencies, 50 ohm home made divider probes have higher impedance then passive probes that plug into a 1Meg scope input. It is not much use offering lots of bandwidth if the loading of the probe is going to change the waveforms anyway.
One of the huge things extra resolution gives you is much better FFT results. A high resolution sampling scope would be great for people who cannot afford or justify a spectrum analyzer.
Last thing - LXI looks like being the new LAN instrumental communications standard. LAN is way better then USB as it has up to 1500V AC isolation, and you do not need a PC on the same desk. You could run the oscilloscope from a smartphone if you wanted using WiFi.
I haven't looked to see if there is a cost, but if you can get you product name on the LXI registered instruments, then your name would get out.
http://www.lxistandard.org/Products/ProductList.aspxRichard.