But naturally my predictable rest of lifetime is not enough for test every possible setup combinations even once per setup. 
If you would choose a 100mHz square, and a 99% duty, you are more likely to hit that high level right at the beginning of the rest of your life.
On that picture it looks like you hit that low level... 
That one magpie's nest type of signal connection that you recently brought up when you criticized the generator's frequency response. I will say one more and one last time. Most of the problems you'll encounter with oscilloscopes or generators are between them and the chair you sit. Next time I might say it less nicely.
Here is 100mHz and 99% Duty. And again hard PowOFF.
And yes, signal is 20Vpp and other is generator 10V DC out (also for trig)
If you set up values, that are far away from those that showed the issue, and want to compare, you also have a problem between the chair and the screen.
YOU ask'd this (if you remember)! And you get it.
And it still didn't show the problem you so desperately hoped for.
Of course it did not mean those values that i told you. 
I talked about spikes that went up to ~5V but you already set a level of 10V! 
Are you trying to hide the issue, or are you trying to make me look bad?
You make a good point here. When the generator is set to maximum output level, the output value is very unlikely to overshoot in the same way you and mortymore have shown. However, with a lower output level, the generator could overshoot when a mains interruption occurs. IOW: rf-loop should test with the same levels you and Mortymore have used. His screendump in the other thread does already show some negative undershoot.
Looks like you did not read what I have told previously. Several tests with also using several different setups, example voltages and so on.
With few exceptions, this forum is starting to look more and more like some kind of clown club.
Take a look example at when @eTobey is wondering about the frequency response (in other thread) and then there is a kind of kindergarten-level connection of the investigated signal...where actually everything is wrong and of course there is a bug or issue in the devices...of course. Here, even vocational school kids don't make magpie nests like that.Hee
Here below some few samples about tests. And these are one shots. Not selected shots like "try until wanted result". Maybe some others can do that on purpose.
SDG1000X
FW:1.01.01.33R1
HW: 02-02-00-24-00 Max Out V: 0-10 Vpk, 20Vpk-pk (HiZ Load and up to 10MHz. Over 10MHz half level)
Source impedance 50 ohm.
2pcs Suhner RG223/U between SDG and oscilloscope.
SDS2504X HD. Inputs 1 Mohm

10 kHz,
1 Voffset, 300 mVpp

1 kHz,
0 V - +5 Vpk
10 kHz,
0 V - +5 Vpk
a. 1 MHz,
17.5 Vpp
b. 1 MHz,
17.5 Vpp 
1 kHz,
17.5 Vpp
10 kHz,
17.5 Vpp
1 kHz,
10 Vpp
10 kHz,
10 Vpp
1 kHz,
5 Vpp
10 kHz,
5 Vpp
1 kHz,
2.5 Vpp
10 kHz,
2.5 Vpp
1 kHz,
1 Vpp
10 kHz,
1 Vpp
1 kHz,
500 mVpp

10 kHz,
500 mVppOf course, the some dog's chin can now wonder why I didn't test the 6 V level, or 3.33V level, is there something I'm hiding. If it continues to happen, I say: Grow up.
I won't publish 300 more pictures because nothing else essential is shown in them.
EOC