I'll note that if the two use-cases are tied together, this basically becomes a minor sub-detail of implementing T293962. (Basically "make sure to implement this in a way that the A/B test can see it", which mostly just constrains that task away from using localstorage or similar.) If we wanted to roll out the A/B test before the hint was implemented (or simultaneously with its release), we'd need to consider those separately.

That said, regarding use case 2, "across wikis" is actually really unusual for us, and I'll have to make sure that GlobalPreferences is acceptable to use for such a thing. (There's a lot of initial-tutorial stuff that you'll see every time you visit a new wiki. There's even a use-case for this remaining so, because the individual communities might want to customize the messages in these things away from a literal translation.)

@DLynch: I think the information you shared in T294167#7460407 contains the missing information we needed to converge on the requirements be ready to implement. As such, I've updated the ===Requirements in the task description.
Please comment here if anything strikes you as unclear/unexpected/etc.

Note: below are the direct responses to the points you raised.

In T294167#7460407, @DLynch wrote:

I'll note that if the two use-cases are tied together, this basically becomes a minor sub-detail of implementing T293962. (Basically "make sure to implement this in a way that the A/B test can see it", which mostly just constrains that task away from using localstorage or similar.)

The details you shared above are helping me to more clearly see the relationship between this work and T293962...thank you for elaborating, David.

If we wanted to roll out the A/B test before the hint was implemented (or simultaneously with its release), we'd need to consider those separately.

We do want to have the flexibility to roll out the A/B Test BEFORE the hint (T293962) is implemented. As such, let's explicitly consider this task as unrelated to T293962.[i]

That said, regarding use case 2, "across wikis" is actually really unusual for us, and I'll have to make sure that GlobalPreferences is acceptable to use for such a thing. (There's a lot of initial-tutorial stuff that you'll see every time you visit a new wiki. There's even a use-case for this remaining so, because the individual communities might want to customize the messages in these things away from a literal translation.)

Noted. For the purpose of this task, I'm going to explicitly EXCLUDE any requirement about the method we use to log peoples' prior use of the New Discussion Tool across wikis.

i. The primary reason for us treating this task and T293962 as distinct is so that we can start gathering data about who's used the New Discussion Tool as soon as possible...I'm glad you (David) raised this consideration in T291536#7424732.

Change 738610 had a related patch set uploaded (by DLynch; author: DLynch):

[mediawiki/extensions/DiscussionTools@master] Set a user preference to record whether the new topic tool has been used

https://gerrit.wikimedia.org/r/738610

Change 738610 merged by jenkins-bot:

[mediawiki/extensions/DiscussionTools@master] Set a user preference to record whether the new topic tool has been used

https://gerrit.wikimedia.org/r/738610

QA: Current state of the preference can be tested via mw.user.options.get( 'discussiontools-newtopictool-opened' ) in the console.

Change 745672 had a related patch set uploaded (by DLynch; author: DLynch):

[mediawiki/extensions/DiscussionTools@master] Update abtest bucketing for newtopictool usage

https://gerrit.wikimedia.org/r/745672

Ignore that, I copied from the wrong tab when writing my commit message.