feat(NODE-7452): restrict server deprioritization on replica sets to overload errors

[tadjik1]

Description

Summary of Changes

• Restrict server deprioritization during retryable operations so that on replica sets, only SystemOverloadedError triggers deprioritization. For non-overload transient errors, the retry goes back to the same server since it was selected for a reason and is still the best candidate.
• New prose tests are implemented.

Second implementation of DRIVERS-3404.

Notes for Reviewers

What is the motivation for this change?

Release Highlight

Release notes highlight

Double check the following

• Lint is passing (npm run check:lint)
• Self-review completed using the steps outlined here
• PR title follows the correct format: type(NODE-xxxx)[!]: description
  • Example: feat(NODE-1234)!: rewriting everything in coffeescript
• Changes are covered by tests
• New TODOs have a related JIRA ticket

[Copilot]

Pull request overview

This pull request implements restricted server deprioritization for retryable operations on replica sets, ensuring that only SystemOverloadedError triggers deprioritization. For sharded clusters, the existing behavior of always deprioritizing failed servers is maintained. This is the second implementation of DRIVERS-3404.

Changes:

• Modified retry logic to restrict server deprioritization on non-sharded topologies to only SystemOverloadedError
• Added prose tests verifying that retryable reads with SystemOverloadedError retry on different servers
• Added prose tests verifying that retryable reads with other retryable errors retry on the same server

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/operations/execute_operation.ts	Added conditional logic to restrict server deprioritization based on topology type and error labels
test/integration/retryable-reads/retryable_reads.spec.prose.test.ts	Added new prose tests for replica set retry behavior with and without SystemOverloadedError

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[dariakp]

Could you copy paste the actual prose test steps on each line as a comment, including any numbering of steps or cases (and make sure the describe and it blocks mirror the prose text as much as possible) - we do this as a matter of best practice when implementing prose tests so that they're easier to audit (and it helps in the rare cases when prose tests in the spec are later amended). You can follow the example of the case above this one.

[dariakp]

Setting a "do not merge" until the spec is merged.

[codeowners-service-app]

Assigned durran for team dbx-node-devs because dariakp is out of office.

Daria's request for changes was to prevent accidentally merging this before the spec landed. The spec has landed and the tests have been updated with the steps from the prose.

[tadjik1]

The base was different and therefore I had to fix merge conflicts with current main.