To me, and my experience with old scopes it looks like you are reducing some information by that dashed line anomaly disappearing through the transformer. I could be wrong but I don't see what else the transformer can really do.
Yes I agree – – I'm familiar with this effect and I am aware that most analog scopes don't show square waves all that well. They do show up often times as dashes. Usually if you play around with a frequency enough, you can get a decent enough view with "legs" however. I didn't even remotely bother to do this during my 10 minute "experiment".
On the other hand, I'm also not looking at or for Square waves. Rather a sine wave coming off of a non HiFi DAC (probably has none of the filtering or AA). When shown on a proper oscilloscope they can go down to the "uV" region, looks more like a stairstep or a set of stairs. Definitely not a square wave, but as you were speculating – – I guess yes it would be shown or manifested as such possibly on this old machine.. And there's no actual "missing" or gaps in information... it's just a nasty looking shape that you wouldn't want to listen to (under the magnifier. Attached is a screen grab of EEV blogs demonstration of this with the 10 X magnifier feature, looking at a DAC based signal generator... (to be fair, this gave me the idea, partially, to even give it all a look-see)
This is where the main argument comes up – – because like you said looking at the big picture it's a beautiful sign wave from either signal source and there's absolutely no way that there's any appreciable missing information. It's all a good point – – but ears hear sound via a different mechanism vs electronic devices and I believe they may be sensitive enough to detect the difference between two identical sine waves (aka the "big picture" ) one being composed of a course signal and one being composed of a fine (a.k.a. analog, never sampled)) signal chain.
Really if you think about it that's all a transformer can really do. I know those dashes looks like the signal is stopping and starting but a transformer could never fill that in.
Now that I mentioned it, harmonic distortion from core saturation at low frequencies could explain a sound difference between the DAC with and without the transformer.
Now you're getting at my overall hunch as to what the transformer is doing. Actually I believe, in one sense, a transformer can " fill in information that is not there" or "missing" ... We would do this is by mechanism of the inductor ( primary coil) ringing of the coil – – adding harmonics ... By this very mechanism would also tend to smooth out the stair steps.. Which of course would be considered "distortion" by most bench geeks (no offense intended here, as we should all strive to be bench "geeks"

. One man's distortion is another man's reverb – – and well ... as we all know, Elvis wouldn't be Elvis without a bit of reverb

.
This would also explain as to why results are so all over the road... depending on the transformer chosen – – and why things can take on such coloration.
Western Electric 91 is a good example of this (and your aforementioned 13k roll-off SE amp) . Pump square waves into it and it looks absolutely atrocious. The ringing of the transformer in an amp with no feedback? Not pretty on the scope.
The amp sounds more than fine however. As a friend into square wave testing has recently taught/reminded me – – a lot of amps even with feedback can get pretty nasty when you pump square waves in. But that's another topic.
If that's what it is, I am OK with that. Because I think analog is the way to go for anti-aliasing – – so to speak. If you can do it with inductors or transformers, it usually sounds better to most people. That method is not popular with manufacturers because it's very expensive. Generally, for most audio engineering / pro audio folks, and those who live in studios – – transformer/inductor based audio processing has been the favored method for eons. EQs, Compressors, Mic Pre's, and also line-level isolation... big arse transformers they love.
I forget what he said this was ... but I think he said 100khz, with 10x magnification. Zoomed out, it's a normal sine wave.
