OK, but you have to set the timebase to something. How do you set it on your scope so as to get the maximum capture depth while also retaining the sample rate you want?
You are overthinking it. Zen master says: 'Try not to find problems that aren't there'.
Except I'm not overthinking it. It's simply a question of how to get the capture characteristics to be what you want them to be. Since there are situations in which you don't care about what the timebase setting is, it follows that you can then use the timebase to get the capture size you want in those situations. Point being that those situations, which you explicitly called out, are precisely the ones where you have the
fewest valid objections to Siglent's "what you see is all you get" approach.
Where the scope behavior you advocate for becomes truly useful is when you
do care about the timebase setting
and you want to use all of the available (or configured) memory for the capture.
All DSOs except for the ones from Lecroy and some older Siglent models, allow at least to force full memory.
You can force full memory in all of the Siglent scopes. It's just a question of how to go about it. For the "what you see is all you get" scopes, you force that via the appropriate timebase setting.
It's a tradeoff. The downside of the "what you see is all you get" approach is that you don't get to use all of the configured memory while having a smaller timebase. The upside of it is that (save for the maximum capture buffer size, which is separately configured) you fully control the entire capture configuration with a single setting (the timebase)
and you always know exactly what you're going to get without having to reference anything other than the timebase settings.
Ideally, you want to be able to choose which approach to take. Fortunately, Siglent has seen the light on that and now makes that possible.
One approach that no scope manufacturer I know of implements would be to allow the user to specify the maximum capture length as a multiple of the time represented by the screen as defined by the timebase setting So, for instance, you might set it to 10x, and thus you'd always capture 10 times the amount of time shown on the screen. Such an approach would retain the primary advantage of "what you see is all you get", namely that it would make it possible for you to always know exactly what you're going to get,
and it would allow you to always zoom out by the configured amount. The remaining memory (if available) would then be used to store prior captures in the way Siglent already does.
And several offer an automatic mode.
It's not clear to me what specific advantages that would have, most especially if the particulars of "automatic mode" are manufacturer-specific, as I'd expect them to be.