I am the one who opened NODE-5421 - which was fixed in 6.2.

I encountered a different (but similar) scenario where the bug still occur.

Use Case

As a developer using the driver
I want BulkWriteResult.insertedIds to always reflect ids of actually inserted documents to the DB,
So that I could confidently rely on the information in this property.

How to reproduce
I will describe one way I found to reproduce it - it is VERY niche - but I believe it reflects the fact that this bug probably occurs upon other kinds of errors - and that fix for NODE-5421 might not be wholesome. So this repro scenario is, I believe, an example that exposes a more general problem (there are probably simpler ways to reproduce):

Assuming you have a `collection` variable which is a db.collection.

1. Define a document to insert copies of:

const documentToInsert = { date: new Proxy(new Date(), {}) };

2. We will need to duplicate/clone this document. `structuredClone` doesn't work (can't clone certain objects), so you need to use another deep cloning function such as lodash deep clone, I use the npm package: https://www.npmjs.com/package/lodash.clonedeep

A function to duplicate the document:

function duplicateDocuments(amount) {
    let count = amount;
    const documents = [];
    while (count-- > 0) {
        // 'clone' is the deep cloning function you use
        documents.push(clone(documentToInsert));
   }
   return documents;
}

async function run() {
    const manyDocuments = duplicateDocuments(10);
    await collection.insertMany(manyDocuments);
}
run();

Result
The bulk insert operation will fail with the error: 
`MongoBulkWriteError: this is not a Date object`
because the date object is wrapped in Proxy
this is acceptable, but `BulkWriteResult.insertedIds` - shows all the ids of the documents, even though zero documents actually inserted to the DB.

Expected result`BulkWriteResult.insertedIds` should reflect actual ids of documents which are inserted to the DB.

I'm attaching screenshots of the error logs.

Acceptance Criteria

Implementation Requirements

• Make the insertedIds map only include actually successfully inserted Ids
  • Should be accomplished by moving insertedIds map construction to after an operation is executed
• Do not impact the performance of map construction.
  • Keep an eye on iterating the array of documents, should not need to repeatedly iterate

Testing Requirements

• Make a local client side error occur when running a bulk operation. Check that insertedIds does not include the failed batch ids
• Check (at least locally) that insertMany performance is not impacted
  • Check an error-ed batch case

Documentation Requirements

• API docs should clarify what insertedIds is set to after an operation succeeds (and after it fails).

Follow Up Requirements

• None