@matmarex might want to correct me, but I think this is a case where the two parses aren't actually redundant. The first parse happens on the text as provided by the user to see whether it's signed (we can't just check for ~~~~ because of templates), then the second parse happens to actually generate the preview if we needed to add a signature ourselves. So anytime this is happening we are parsing different wikitext.

@Ladsgroup I looked where the error's triggered in ParserObserver, and it happens any time there's multiple calls in a single pageload to parse a page with the same title, revision, and options. A reasonable thing to do here might be to suppress this error if the parse-request provided its own wikitext rather than fetching the stored wikitext for that revisionid, since I think the error is much more targeted at "you are repeatedly fetching-and-parsing a stored page"? Or perhaps include a hash of the provided-wikitext in the options hash?

Thanks. I don't know the details of DT but it would be great to avoid calling parser twice if it can be avoided and if it's not, there should be a way to avoid this logspam because it's drowning the actual problematic cases.

Change 791702 had a related patch set uploaded (by Bartosz Dziewoński; author: Bartosz Dziewoński):

[mediawiki/core@master] ParserObserver: Only report duplicate parse if the content is the same

https://gerrit.wikimedia.org/r/791702

Change 791702 merged by jenkins-bot:

[mediawiki/core@master] ParserObserver: Only report duplicate parse if the content is the same

https://gerrit.wikimedia.org/r/791702

@Ladsgroup Thanks for the review, sorry it took me so long to get to this.

Moving to blocked, we should check the logs after the change is deployed to confirm the issue is fixed.

SGTM

It looks fixed.