A reaction crossing between platforms shows up in the thread as a written description of the emoji rather than the emoji [2], and a reply arrives without the message it was answering attached [1]. Those are small artifacts. They are also why a fair amount of messaging code exists.
This is for the person who owns a branch in the send path labelled something like "iPhone recipient, degrade gracefully." That branch usually holds two different kinds of thing. One is visible: custom strings, a help-centre article explaining why reactions look strange in mixed threads, a nudge to move the conversation into your own app. The other is invisible logic nobody has read since it was written: SMS length assumptions, reaction suppression, reply context stripped before send. iOS 27 makes the first kind embarrassing and the second kind deletable.
What is being pitched as parity is really the removal of a branch, and that removal has a schedule problem. 9to5Google's tracker is a useful corrective to reading "rolling out" as "shipped": it notes that Messages features are A/B tested and that the RCS client can take a long time to move anything into stable [4]. Take its own example. The first read receipts redesign started appearing in August 2024 and is only now seeing wide beta and stable availability [6]; the tracker entry is dated 29 August 2026 [8], which puts roughly 24 months between first sighting and broad availability [7]. The interop items do complete, though. Encrypted RCS between Android and iPhone arrived with iOS 26.5 and now sits on the stable list [5]. On the Android side the client you test against is narrowing, since US Samsung users are being moved onto Google Messages, which lands in the homescreen dock automatically on Android 14+ [9].
What the tracker cannot tell you is how many of your users are in mixed threads, or whether the degraded reaction ever cost you a conversation. It tracks releases, not how people are actually using them. What it dates is the retirement of a specific excuse.
So sort every mixed-thread assumption you hold by what its removal waits on. Bucket one waits on your own code: the copy, the help article, the nudge. You can schedule that. Bucket two waits on Apple's version adoption, which means a version check plus an adoption threshold you pick deliberately and write down, rather than a vibe about when "most people" have updated. Bucket three waits on Google shipping the receiving half in stable, which means somebody verifies it on a real device before the branch comes out. Only bucket one is yours, and it is usually the bucket teams leave alone longest because deleting an explanation feels like less work than it is.
The check on whether you got it right is not how many messages move through the thread. It is the fallback branch firing less often in your logs, and support closing out the ticket that starts with a sentence where an emoji should have been. Parity shows up first as a version floor. That floor, and not any announcement, is the unit worth planning against.
