A few days ago someone removed the en-gb label (which was identical to the en label) from P973 (described at URL).

When I searched for "desc", it no longer appeared in the first set of results:

After I undid the edit, it reappeared:

(if it matters: these screenshots were taken while adding a new statement on https://www.wikidata.org/wiki/Q13273696)

Here's the results for "desc" on the same item with the interface language set to en:

and en-ca:

There seems to be a bigger issue with ranking, because those results are worse than en-gb.

en-gb:

en-ca:

en:

I would say the best two matches for "desc" are clearly "described by source" and "described at URL", because they have a lot of statements in general, have been used on items with the same P31 and have labels starting with the same string that is being searched for, closely followed by "main subject" where it's matching the beginning of an alias rather than the label.

It makes no sense at all for "taxon author" and "year of publication of scientific name for taxon" to be so highly ranked. They're not used as main statements, this isn't a taxon item, and the search string doesn't even appear anywhere in the current label, let alone at the beginning.

FYI, I was the one who removed the (ostensibly) redundant en-gb label @Nikki spoke of. I find this complication to be emblematic of a more broader issue—the need to distinguish regional variants from an ultimately "neutral" variant (which in English's case, is unfairly assumed to mean en-us). It would be nice if regional variations were defined explicitly, perhaps laid out like this:

I'm aware this issue is partially addressed by T303677 (as many of these variations tend to be duplicates of each other), but this approach deals with the issue of locale-sensitive translation in a culturally-neutral manner (as well as making variants more visible to editors, who may neglect to update regional spellings when modifying descriptions and labels).

Hopefully this makes sense.

In T334563#10054830, @Alhadis wrote:

FYI, I was the one who removed the (ostensibly) redundant en-gb label @Nikki spoke of. I find this complication to be emblematic of a more broader issue—the need to distinguish regional variants from an ultimately "neutral" variant (which in English's case, is unfairly assumed to mean en-us). It would be nice if regional variations were defined explicitly, perhaps laid out like this:

That looks almost exactly like what I proposed ages ago in T209068 :)

(There's also an even older ticket about making en more international - T33874)

I'm aware this issue is partially addressed by T303677 (as many of these variations tend to be duplicates of each other), but this approach deals with the issue of locale-sensitive translation in a culturally-neutral manner (as well as making variants more visible to editors, who may neglect to update regional spellings when modifying descriptions and labels).

Removing descriptions should be fine, that doesn't seem to affect the search ranking.