> db.createCollection("special", {collation: { locale: "en", strength: 1 }})

> db.special.explain().find({_id: "123"}, null, { collation: {locale: "simple" }})
<snip …>
winningPlan … COLLSCAN

The above query doesn’t use the collection’s _id index because of a locale mismatch. In the case of a locale:simple query, though, the matched documents will always be a subset of those that would match with the _id index. It’s thus possible to match along the _id index then apply a binary filter.

Currently this is possible with the following workaround:

db.special.explain().find({
  $and: [
    {_id: "123"},
    {$expr: {
      $eq: [
        {$_internalKeyStringValue: { input: "$_id" }},
        {$_internalKeyStringValue: { input: {$literal: "123" }}},
      ],
    }},
  ],
})

… but the above relies on an undocumented operator and is quite a hack besides.

The hack also doesn’t work with upserts since upserts forbid $expr in the filter.