I'll note that the article you cite explains why customer frustration should be avoided, but doesn't really cover the topic of modal dialogs. A better source might be something like Nielsen's Modal & Nonmodal Dialogs article (especially the "disadvantages" section), which concludes that "modal dialogs become problematic when used for noncritical activities." Mind you, Nielsen's article doesn't provide citation (though perhaps Nielsen's name recognition is enough for it to be convincing).

I'm a big fan of the "clickable premium button" approach. On the tagging side, typically I've seen people use a dollar icon rather than a lock icon.

Are users pushing both buttons because they're confused or because calling two elevators reduces their expected wait time?

I'll note that this approach lends itself very well to users typing in the number by hand (round to the nearest valid value after the spinner loses focus). While this is still possible with a combo box using Roger's approach, a slider allows a user to jump between adjacent values when the rounding is in the wrong direction. Whether this advantage is relevant depends your use case. I could see it being relevant if the user's motivation is, "I want the number to be as close to X as possible. It must not go over (under).

I'll note that "technical limitations" is a poor reason to avoid using a slider. It's not hard to modify a slider to use inconsistent step. Mind you, you this leads to the UX choice that slider positions are either not proportional to value or the slider steps are uneven. Since you mention technical concerns, I'll note that the former is trivial to implement using off-the-shelf components whereas the latter is not. For reference, musefan's answer uses the former approach.

It's also probably harder to tell apart quickly. If you're making a time-based color game, showing 4 and 7 on the same screen is probably going to add an unintended challenge.

I'll note that you can, with decent (but not perfect) reliability, detect if a user is a current user. E.g., leave a "this is a logged out user" cookie behind when users log out.

I'll note that you can, with decent (but not perfect) reliability, detect if a user is a current user. E.g., leave a "this is a logged out user" cookie behind when users log out.

To point out the obvious: Before asking the author for guidance, check if the author has already provided guidance. Such guidance comes in two forms: 1) The author asks you to do something. 2) The author's license requires you to do something. For example, some freeware software explicitly forbids hosting the download yourself, whereas GPL software always allows you to do so. After all this, you'll know, "I'm allowed to do X,Y, and Z. The author's preference is for Y and Z." Mind you, none of what I said is UX-related.

@IlariKajaste: I'm not so sure a small fine would be OK. Freakonomics found that, in at least one case, small fines run the risk of causing an increase in poor behavior. You've gone from "forcing the city to waste resources rescuing me is bad" to "the city charges a nominal fee for rescues."

Why not just add aria-haspopup="true" to your top-level items?