Dev Feed

Seems like a general AWS SDK issue in Lambda only. We get this for SecretsManager SDK. AI tooling recommends moving the instantiation inside the execution context, but that hasn't worked, which is very weird - that signals some underlying cache behaviour that we can't control outside of the execution context? None of our latency logs point to any kind of delay even close to 5 minutes: I'm working on high throughput, fast acting lambdas by design ~10-100ms. This is contributing about ~4,000 errors a week across our ecosystem out the moment. A real noise issue.

Same issue here with nocks that previously worked (v13) to intercept requests to AWS e.g. something pushing to S3. They fail on Nock 14 with a `NetConnectNotAllowedError: Nock: Disallowed net connect for...bucket.url` error. We only have `enableNetConnect(127.0.0.1)` for our suite to completely isolate all tests, but given we nock the AWS URLs I wouldn't expect this to hard error (and previously, it didn't).

Interesting point on OS. I'm running locally on `macOS`, but in the cloud on `ubuntu-latest`. Just had this come up again Github Action <img width="792" height="102" alt="Image" src="https://github.com/user-attachments/assets/552f5d4c-9f6e-4ba5-9420-495c2bdd1104" /> Local <img width="1324" height="69" alt="Image" src="https://github.com/user-attachments/assets/0888d130-0799-47ee-a53c-1526e4ccf863" />

Release - Published

 Still not using prettier because of this. Time to unsubscribe from this thread 😃

To be tested elsewhere before raising a separate PR