To me convoluted process is to set memory depth, mentally calculating what timescale it will be at target sample rate, mentally calculate how much of that I have to set as pretrigger to make sure I have enough pre trigger capture once I'm at the target timebase. Then I set target timebase, tightly set around trigger and then keep capturing until I find something interesting, and then, without any sense where exactly in the whole buffer I am i keep twiddling around with timebase looking at data searching for something interesting.
Problem is that Dave misunderstood (or ignored) original reason why Nico is such a proponent of this.
Nico uses it for slow, long samples, where he looks at long bursts of data on ,say, CAN bus. He keeps looking at one packet or event, and when that even happens he wants to know what happened around it. He doesn't stop captures, but works in RUN/Normal triggering mode (not Auto), from trigger to trigger event, and he initiates data transfers from equipment manually, or has them set up to occur slow enough that he has time to manually stop scope if needed.
It is very deliberate, he doesn't mind that scope needs 1 second per capture to process, and it IS NOT interactive scope work. You cannot have interactive scope work with 2 triggers per second.
As I already documented, Keysight scope DOESN'T work like that (Nico uses his R&S RTM3000 for this) in RUN mode. In RUN mode in between triggers, it has only screenfull of data available. Only after you STOP, it will gather all buffers and reassemble them. And if you are lucky that you are working in less 5 us/div, you will have data up to 20 us/div. If you are at 20 us/div and up (50us/div in single mode), there is NO data outside...
I had DS1074Z. I pretty much never used fixed memory length. I used AUTO mode pretty much exclusively. Setting scope to capture too long capture on every trigger makes scope slow. Every one. One of the reasons Tektronix get their bad reputation is manual memory settings. People put them to 20MS and then complaint how it non responsive.
Suddenly that is a good thing. It is not. That being said I DID USE IT from time to time. For instance to look at gated PSU startup. Set it for MAX mem, set timebase for switch on event, look at it and then if there was something wrong, pull out to see what was happening around it. But with Picoscope's excellent zoom implementation, it behaves exactly the same the other way around.
Siglent 2000X+ seems to work exactly as Picoscope. And it is not "badly implemented manual memory length", but a "auto memory management with maximum length" setting. It is there to control strategy to balance huge memory with how will sampling rate drop off as you go to slower timebases. Siglent should have better explained and/or named this option. Siglent doesn't have "fixed manual memory size"
Also, if it is possible, they should think of implementing Keysight like feature in user configurable option, that they give up history mode and do single shot captures in full length as set. That would make them almost exactly same as Keysight (same at slow timebases). They might even do "full Keysight" and set minimum sampled length. Whatever.
That being said, if customers think this is necessary, manufacturers should try to address it.