In order for some observation attributes to be searchable (especially attributes inhered from associations like with taxa, places, projects, etc), observations need to be re-indexed in our search database. This re-indexing usually happens in a queue, so when the indexing happens is variable depending on how filled the queue is.
In this case to determine which observations were affecting the results, I clicked your link to the observation search species count page, then clicked the “X observation” link under that taxon. That narrows down your search to filter by that taxon. Then for each I used a trick to manually re-index them which is to fave then un-fave the observation, which triggers an immediate re-index of that observation. I did that for 4 observations and I think that resolved the problem. Keep in mind that search results for a given query can be cached for 5 minutes, so in order to be sure the re-indexing had an effect, you’ll need to add a search parameter (I usually add something I know is true for the observation like the place, date range, or user ID)
I’m not 100% sure if the re-indexing would have happened eventually - it is possible there’s a bug with check lists that the re-indexing jobs aren’t created properly. If this happens again and doesn’t clear up after about a day, then there’s likely an indexing oversight which would be good to identify.