This feature (parsing only stable version of templates when parsing the articles) is extremely taxing on the infrastructure. It requires a table that usually ends up being the biggest table of that wiki (by a wide margin, e.g. arwiki one was bigger than revision table of English Wikipedia) and has contributed to the general outage of March 6, 2024. It adds a lot of writes to the every edit and given that FlaggedRevs is unmaintained (T185664), I don't see an easy way to fix this.
On top, this feature is not that critical. Many wikis rely on protecting templates instead (as there is no reason usually for an IP/newbie to edit them). Even so, it could be theoretically re-implemented in such a way that wouldn't be so taxing on the infra (e.g. instead of keeping track template revisions, just get the latest stable revision) .
Proposal: It's currently on by default, notify the communities that default will change and they can ask to be a holdout if they heavily rely on it. After a week or two, flip the default and clean up the tables from production. If there is no holdout there, just drop the feature/code altogether.