tldr: this may actually boil down to a Parsoid issue.

If we look at the Parsoid version of the talk page in question (which is used by DiscussionTools API to give us the Talk messages and coffee-roll headings):
https://id.wikipedia.org/api/rest_v1/page/html/Pembicaraan:Indonesia

...and now look at the "source" of that page, we see a couple of places with slightly unusual markup:

...
data-mw='{"attribs":[[{"txt":"style"},{"html":"\n    text-align:center;\n    white-space:nowrap;\n    font-weight:bold;\n    background:<span typeof=\"mw:Nowiki\" data-parsoid=\"{}\">\n#b2ff66</span>;\n    "}]]}'
...

If you squint your eyes, you can see a weird bit of markup near the end: the closing span tag begins with a &lt; entity, but ends with a literal > character, which is inconsistent. It should be either <span> or &lt;/span&gt;.
Technically none of this should matter, since all of that markup is part of a data-mw attribute contained in a string, but I'm guessing for whatever reason your (iOS') HTML-to-native text parser might be tripping up on this inconsistency, while our Android parser simply ignores it. This is just a guess, but it seems to be plausible because the weird markup is rather close to the "unparsed" text seen in the screenshots.

Android parser code in case we want to try to compare missing stripping logic: https://github.com/wikimedia/apps-android-wikipedia/blob/660f9de29d5f4cd2a35ae34339c24c08cfb72044/app/src/main/java/org/wikipedia/richtext/CustomHtmlParser.kt#L117