[Alex KL-1]

Excellent question.

I connected other sources, and effect are still there, only for the interlaced ones.... for the Atari, the issue simply vanished.

For completness, I connected the supposedly worst offender (old notebook with a S-Video out) to a different old TV. Nothing similar. The image lines are straigh, but then I noticed when it's loses interlace (oh, that subject...), the effect resembles it a little.

Then, I plug back in the problematic TV, and I noticed the problem almost vanished! I noticed I needed to readjust the V. Hold for one source (the Atari). I proceed to adjust the vertical hold, height and linearity, and I'm able to "move out" the effect. After some tweaking, I ironed out the bad effect, making it invisible.

To be sure, I subbed the inverter back to a 6CG7 (I see I subbed it for a ECC88, for raising the pulse amplitude for the vertical sync), and no changes in the H; only the V are affected (changed the sync locking point).

At first I'm pretty sure this effect being from some H issue, due to visual effect, but the vertical maybe is triggering to a strange portions of the sync (or suchlike), since:

This TV is a Admiral that have various issues (badly oxided tuner, missing cores for some IF coil, bad flyback (secondary), bad Vout transformer), and then I decided to consider a disaster as an opportunity, modding it, for fun (through learning). And I like to make some circuits of my own (generally for audio).
The H circuit is original, with it's sync inverter. The vertical is heavily modded, with interlace diode, Miller oscillator, and negative current feedback output. This is working very well (the far best old TV I have for interlace, rivalling with the one-chip modern TV), but it appears to have some side effect, then (even when image was interlacing rather well). The interlace then results textbook perfect when I move out the bad effect. Or, in other words, even with interlaced diode, I cannot choose an arbitary sync point; is needed to stay under certain range.

Perhaps I need to check the interlace diode bias...

(for completness, the mod video amp for this TV I inspired from the Conrac monitors, with 3 stages heavily loaded for wide bandwidth, since the modern devices have only 1Vpp instead the ~4Vpp from video detectors from this TV time, and only one pentode not suffice for gain. Difficult was to enforce a good response without having a wideband oscilloscope... but it seems to be ok; it not have too much ringing or other artifacts, and resolution is second to none)

Besides the "disaster" and my hobby of making circuits, one inspiration is the TV kits of old times, for building it at home, and my 29" TV with TDA9373 (that I asembled of my own, fully, minus the TDA programmation; is from a commercial TV).