1. Is it possible to switch off the sinx/x interpolation to see the raw sampled data points on the display?
See the chapter on "Display Control" in the manual. In the online copy I found, it's the first section of chapter 16, "Display Type".
It it usual that the oscilloscopes' cursors work that way or it is a MSO5000 specific?
It's certainly the standard behaviour of oscilloscope cursors to work in the screen region only; I am not sure whether some brands offer alternative modes. If you can use an automated measurement instead of cursor readout, that will typically take a lot of uncertainty away.
To precisely measure time delays between two events which you need to select manually, reading the delay time instead of using cursors may be more precise: Zoom in (on the time axis) onto the first event, position it e.g. on the central vertical grid line via the horizontal delay control, and take note of the Delay readout on the screen. (I.e. the time delay from trigger event to centered event.) Then repeat the same with the second event and take the difference of the two Delay times.
I noticed that the software update rate of this scope quite slow. Worth to communicate the problem towards Rigol, do they usually care of this inquries of the users?
You mean the response to user interactions like turning the position knobs? I'm afraid what we get there is the best the CPU can do. Compared to other DSOs it is "par for the course" -- not something to brag about, but some competiting scopes are doing worse.
Rigol is certainly aware of the limitations; but given how long the MSO5000 has been on the market, I would not expect that they can (and/or want to) improve this.