Following SERVER-113685 / AF-8199 / CVE-197 , a request was made to audit our testing coverage for _id. There are two things that are _id:

• the _id field in a collection, which can be used in various DDL admin statements and such
• the _id field in a pipeline, which may come from a collection or be produced by some pipeline stage.

The test coverage for both uses of _id was audited. While there is a lot of use of _id and some incidental coverage from unrelated tests, the following cases do not seem to be covered completely or systematically:

1. The use of _id in createIndex() in jstests is somewhat incidental, and does not cover the full matrix of possible cases:

_id with different field arrangements and sort orders
_id with different index types
_id with different index options (collation, clustered, unique)

2. There is no systematic coverage for_id being redefined by the various pipeline stages, such as $addFields, $replaceRoot, $unwind, $lookup, $merge, various aggregations ($densify, $fill, etc) followed by an attempt to use said redefinition of _id

3. Usage of _id in a DML MQL, such as findOneAndDelete, bulkWrite

4. Use of _id in expressions, such as $toString , regex

5. Use of _id in schema validation

6. Use of corner case _id values are rejected (MinKey, MaxKey, maximum size, etc)