If you do not feel the purpose of this issue matches one of the types, you may apply the typeignore label to exclude it from type tracking metrics and future prompts.
This groupproduct planning bug has at most 50% of the SLO duration remaining and is an SLONear Miss breach. Please consider taking action before this becomes an SLOMissed in -1013 days (2020-06-08).
I did some network request inspection here, and for gitlab-org/gitlab project, we're fetching all the open issues with their iid, title, and id, over 50k issues that is, and thus, request takes ~8 seconds and downloads about 4.9 MB of payload. However, once this list is loaded, it is cached as long as page is not reload so successive searches are faster because the frontend does it locally. which IMO is quite fast for the size of data we're fetching here.
The endpoint in question here is /gitlab-org/gitlab/-/autocomplete_sources/issues and it doesn't support page size or page number IIRC, so there are two possible approaches we can take to fix it this;
Backend + Frontend
Update /autocomplete_sources/* endpoints to support params like page_size, page (i.e. number), search (that accepts characters typed).
Only drawback here is that every key-stroke would cause a network request (unless it was already made and cached), so depending on network conditions, it can be fast or slow.
Frontend only
If we know usage for # is high, then we can proactively fetch the response and cache in the browser (either in localStorage or IndexedDB), and perhaps periodically refresh it. This would ensure that autocompletion results are loaded instantly. This can be done as an extension of #386275.
Obvious drawback here is the lack of real-time data, where the user may not have a newly created issue showing up in autocompletion and would have to wait for background refresh to complete.