Modern digital scopes are no different than analog scopes. The sample rate (along with SinX/X interpolation) is high enough not to cause the problem you are alluding to.
So a digital scopes bandwidth roll off will basically be similar to that of an analog scope, as will be the trigger system as well.
That's what I used to think, but according to this post and video, the roll-off of digital scope is MUCH steeper.
https://www.eevblog.com/forum/testgear/digital-oscilloscope-behind-the-scenes-presentation/
What bandwith you need depends on the signals you want to look at.
The 5*fc is very roughly a rule of dumb for square wave type signals, simply because of harmonics.
Although a square wave with only 5 harmonics still doesn't really look like a square wave.
You'd rather need the 9th harmonic or so.
They used to teach us that the 7th harmonic was the bare minimum to get a reasonable square wave.
That's correct, the bare minimum
So you want to have just a little more than that.
Consider a 1000Hz signal coming from a modern logic gate (e.g. 74lvc1g*) with a rise/falltime of ~1ns. You want to see whether the signal integrity is valid. What bandwidth do you need? Hint: it isn't 7kHz!
If you think that example isn't relevant to RF, then consider a cheap single-chip 4.4GHz frequency synthesiser with a VCO operating from 2.4-4.4GHz. If you tell it to generate a 40MHz output, that will be done by a digital divider, so the output will be a square-ish wave with edges that are sufficiently fast to generate a 4.4GHz sine wave.
More anthorpomorphically, consider that when displaying a transition, the scope neither "knows nor cares" whether the next transition occurs in 100ns, 1us, 1ms, 1s, ...
Summary: who cares about harmonics of boring square waves; all that matters is the transition time.