With an ever-declining IE/Edge market share, I usually rely on the principle of progressive enhancement and accept that certain non-essential features will only be available to users of modern web browsers (unless, of course, a specific project has an unusually high number of IE/Edge users, e.g. intranet applications).

Declining, but still non-trivial.

We dropped IE10 support when we stopped doing IE9 a year or so ago because they both had similarly insignificant numbers, and the opportunity cost of supporting them was too high.

IE11/Edge are also showing fairly similar numbers, which suggests edge uptake is slow, but even factoring in the extra hours, the bugs they generate (usually weird flex issues in both cases) we still effectively make more money by supporting them than not.

Yeah, relying on progressive enhancement is obviously only viable when we talk about nice-to-have features. Sticky headers will often fall into this category.