On the scopes I have worked with the number of channels you choose to display does not change the bandwidth of the channels.
In general for digital signals you would like to bandwidth of the scope to be 10 to 20 times than the data rate. It depends a lot on what you are trying to look at with the scope.
Yes, it’s the 3 dB corner freq of the scope’s input section. In practical terms, think of it this way. If you put a super-fast rise time signal into it, the scope will display that signal with a rise time of (0.35/BW), or in a 100 MHz case, 3.5 ns. So if your digital signal has a pulse width of 10 ns, it will be displayed with best-case rise and fall times of 3.5 ns.
If the input signal is not way quicker than the scope, then the displayed rise time will be the root-sum-square of the scope’s own rise time, and the signal’s rise time. If your signal actually has a 3.5 ns rise time, the scope will slow it down further so that it displays 5 ns.
As others have said, the bandwidth is the LP cutoff frequency (-3 dB point) of the oscilloscope’s analog input stage. A good rule of thumb is that the oscilloscope’s bandwidth should be at least 5X the highest frequency component you want to measure. This rule of thumb will give you an uncertainty of around ± 2% or better.
Be careful with digital scopes. A lot of times they will advertise a 100 MHz scope and what they really mean is that it is a 100 M s/S scope, which means the usable bandwidth will be much less.
The fastest signal you can see can be higher in frequency than the 3 dB upper corner frequency. However, you won’t be able to make accurate amplitude measurements using the display on the screen. That is because the amplitude response of the amplifiers falls off quite rapidly above the corner.
For looking at square waves, the fastest rise time (to 63% of full amplitude as I recall) that a scope with 100 mHz. upper corner can display is about 1.6 nanosec.