What stage of CSS specifications do browsers base their vendor prefixes and experimental flags on for CSS properties?

I understood that while a W3C recommendation for a feature was still under development, it would often already have reference-implementations in browsers, but behind a vendor prefix or an experimental flag.

So it would be very clear in code that you're relying on a standard that might change.

Today I realised that this was not at all true. I checked my favourite example, position: sticky, and it seems that Firefox removed its experimental flag 2014, Chrome in 2017, and Edge directly introduced the feature un-prefixed in 2017.

Now the related standard CSS Positioned Layout Level 3 is still a Working Draft, implying there might still be changes until the official recommendation.

So I'm left with two questions:

1. On what indicator do browser vendors base their decision to remove vendor prefix or experimental flag?
2. Is there some formal state within the drafts that tells them (and us) that a specification can be considered stable enough for implementation?

It's mostly the other way around. A W3C spec can only become a recommendation when it has two complete, independent implementations. Officially, at least. Sometimes the rules are bent a bit.

It matters less than you think. Once there are interoperable implementations for most of the browsers in current use, any variation of the specification from the implementations is normally considered as a bug in the specification, not in the implementations. So at that point the spec is effectively stable for that feature, regardless of its status as Draft, CR, PR, or Recommendation.

Browser makers talk between themselves and the spec writers to decide when they're going to implement something, and when they agree that their implementations are complete. The "Recommendation" status is just the final step, to tell web authors it's all good for them to use it.

Of course, only the most defensive of web authors will wait that long.