If it behaves the same at 1Vpp where the spec says it should be <10ns then it's an issue for sure. I just tested this on my DG4000 and expected the same result but its edges looks the same at 10s period as they do at 10us - for whatever that's worth. On the DG4000 the square waveform is "better" than the pulse waveform when it comes to edge times at lower frequencies.
Yes, there is indeed a problem there with Pulse, even at 1Vpp..
I will show same with SDG1032X, attached images.
Pulse freq 20mHz 1Vpp
With 50% duty and also set for 50ns width and same 20mHz frequency (50s period)
Just the usual tradeoffs between performance, samples, frequency, jitters and so on in a digital system. This is most likely the worst case, and on a smoother waveform would most likely be the correct thing to do..
This looks more like result of poor desig than "usual tradeoff...etc"
I wonder how there can fine adjust pulse rise and fall times and pulse width with these long period times pulses when transitions are like in these examples.
------------
In my example (in previous msg and now also here) there is just normal simple cheap DDS Arb function generator. (SDG1032X)
In first image pulse period 50s (f=20mHz)
Second image same 50s period but pulse width adjusted for 50ns. (and width and rise/fall times can fine adjust <1ns resolution independent of pulse period, even if period is 1000000s. (factory versions are limited for min 32.6ns width and min 16.8ns rise and fall. This individual unit Pulse rise and fall minimums are mod.)
Square rise and fall is more fast, around under 4ns. (because these can up to 60MHz square)