Across various makes/models I think there will definitely be variation in the amount of ESR at the input connector. My Tek 465 is spec'd at about 500Vpkpk (at VLF) and it also states the input impedance is 1Meg in parallel with 20pF.
So based on this it ought to be fairly bulletproof. But in reality there is a series 51R damping resistor at the CH1 and CH2 input. So if you look at the impedance of 51R in series with about | 20pF + 1Meg | in parallel and do this from LF to 100MHz the input impedance changes a lot over frequency.
At RF the 1Meg resistor can be ignored and the input impedance will be dominated by about 40 ohms ESR in series with about 20pF. I think some of the 51R (maybe 10-15R?) gets cancelled by the negative resistance of the active follower and this typically leaves about 40R in series with 20pF.
My spreadsheet predicts the parallel equivalent of this at 30MHz is about 1600R in parallel with 20pF. So the 1Meg Rp shown on the Tek 465 spec sheet is shot to hell by 30MHz
I think my HP 54528A scope is rated at 5Vrms with the 50R mode selected but I don't know what frequency range this applies over. This is about 0.5W. My previous scope was an old 500MHz HP54540C and this had a very accurate internal 50R termination but I don't think it was accurate over the full 500MHz of BW. It was very good up to about 150MHz I recall. I'd expect the 54825A to be very similar as I suspect some of the acquisition board technology in this scope will be similar to the older 54540C. So I might have to derate the 5Vrms spec slightly if I wanted to use the scope up at several hundred MHz with the internal 50R load selected.