There's probe bandwidth and there's channel bandwidth.
Probe BW is lower as far as I've seen (because what's
the use of more BW than the 'scope front end can pass?).
500pS risetime is (if you call it a quarter cycle) 500MHz
(-ish) and you can find passive probes of about this range.
The idea of using 50-ohm cable and a well constructed
passive divider is a good approach provided that your
source impedance (hence acceptable divider impedance)
is low enough to make sense. 1Mohm against a 50 ohm driver
is wasting BW. I doubt you're seeing 500pS risetimes in a
high-Z system.
- - - Updated - - -
Another thing - it may not be probe BW per se, but probe C
and source Z combining to simply degrade the edge rate.
I doubt you'll see LVDS, PECL, CML rated for transmission
lines with excess C - they rate, and design to, as little shunt
C in the "load" as they can (or their customers let them) get
away with. 10pF and 50 ohms, even, there's your 500pS
even if probe backend and channel were infinite BW.