doc: announce upcoming end of tier 2 support for macOS x64

[aduh95]

The timeline will hopefully be (unfortunately order is not guaranteed):

1. 2028-04-30: Node.js 24 EOL (Node.js 24 lists macOS x64 as Tier 1)
2. Early 2028: MacStadium end of support for our testing infra
3. 2029-04-30: Node.js 26 EOL (Node.js 26 lists macOS x64 as Tier 2)
4. 2029: Apple end of support for our building infra¹

Refs: nodejs/build#4317 (comment)

Footnotes

1. https://developer.apple.com/documentation/apple-silicon/about-the-rosetta-translation-environment/ announced that macOS 27 will be the last version with full Rosetta 2 support, which we use to build the universal binaries. macOS 27 EOL is unknown, based on previous version we should expect it to last until 2029 (and can get away with running it post its EOL as those machines only run trusted code). ↩

[nodejs-github-bot]

Review requested:

• @nodejs/build
• @nodejs/tsc

[richardlau]

FWIW this isn't really an announcement -- I don't think people in general will notice this change unless attention is brought to it (either they are looking when upgrading from older versions or someone has pointed them at it).

[mcollina]

lgtm, we should also include a blog post.

Note that by this timeline we would have 2 years for Node 26 without the testing infrastructure.

I don't think this is tenable, so we'd have to lower it in flight for Node 26.

[aduh95]

@mcollina only one year. This PR goal is to document that macOS x64 will be de facto experimental as of early 2028.

[Renegade334]

+1 for a blog post.

[sxa]

Can I just clarify something before this is approved:

will no longer be able to test changes on any Intel-based macOS version, so the only difference with Experimental tier is that the project intends to keep creating universal binaries, compatible with both Apple Silicon-based and Intel-based macOS versions.

Is the intention here that we would be shipping the universal binaries after MacStadium shuts down the x64 hardware meaning that the x64 side would be completely untested despite being distributed through the formal channels?

In reality "The only difference" is therefore that it is completely untested and not in any way guaranteed to work in any capacity - is that correct? If so I'm not sure "The only difference" in the current sentence specifies any difference ... Either that or I'm typing this so late that I've missed something obvious but I'm going to "request changes" to stop this being merged until that is clarified :-)

(Also noting that we would technically have the option to test under Rosetta on arm hardware for a little longer while that is available if desired)