"never ask me again" is hard because it is essentially a setting, and if you introduce a setting, you need a way to turn it back off, including the design, UX work, localization work, testing work etc. that is involved in adding a new switch. If you get to 1000 of these, you probably need categorization and search. Oh, and you need to track which ones are even relevant for that user; it doesn't make sense to ask a German to link their Comcast account. You need to make it all work consistently across platforms, except for the settings that should only work on one platform. You get my point.
"remind me later" is simple; you check if there are any children accounts, if no, ask to create one with probability p. There is no state. There is no setting. There is no "what if I accidentally clicked no but want to reverse that decision, where do I go" problem.
As someone who’s worked on dark patterns like this in the past, I can assure you, difficulty of creating a new setting is not the reason for the “show this less” pattern.
It’s much more simple: if a “permanently off” setting is worse for metrics, it won’t get built that way.
All of these elements are table stakes for software that people will use and application settings is not a “hard” problem.
Normally it would just be lazy design to annoy the shit out of your users with nags instead of giving them settings but this approach seems to be pervasive at companies that make billions off this software so I can only conclude that it’s intentional.
This means that there is actually likely MORE state tracking this than if it were a simple setting. You need to store not just the setting for this particular feature, but also potentially the last time they were asked or whatever else triggers them to get asked again. It’s either poor design, or explicitly designed to repeatedly push a feature on a user even if they don’t want it.
"remind me later" is simple; you check if there are any children accounts, if no, ask to create one with probability p. There is no state. There is no setting. There is no "what if I accidentally clicked no but want to reverse that decision, where do I go" problem.