One thing to remember here is that a typical 10X scope probe is not a complete 10 to 1 divider. It is only a 9 Meg series resistor. It depends on the 1 Meg to ground impedance of the scope input to complete the divider.
For any downstream attenuation devices to be accurate when inserted between a scope probe and the scope they must:
1) have and input impedance of 1 Meg, and
2) be characterized to work into the 1 Meg input impedance of the scope.
That is true but it is the capacitance that causes the problem when connecting attenuators, including the normal 10:1 probes, to scope high impedance inputs.
A good scope has a consistent input resistance and capacitance of 1M Ohm in parallel with 20p - a surprisingly large capacitance - and that is why 10:1 probes have frequency response adjusters to compensate for the scopes input capacitance. The probes output capacitance, mainly caused by the probe's lead, is another variable.
On a good scope, the input capacitance will be constant for different Y gains, with frequency, and with signal amplitude variations (setting up a scopes input capacitance is part of the calibration procedure on a good scope).
On a not so good scope this is not the case and while this may not be a problem when measuring a low impedance signal without a 10:1 probe, it can introduce errors when a probe is used to measure high frequency signals.
Of course, if you are measuring low frequencies the capacitances have little effect.
It is quite interesting to see the difference in display between a scopes 50 Ohm input and the 1 M Ohm input, not to mention the difference when a 10:1 probe is fittted to the scope.
spec