Proposed changes

1. Typo fix closes [docs] deleteBeforeReplace affects dependents #13934
2. Additional warning for eventually consistent systems.

For (2), we recently ran into this with AWS. In this scenario, an IAM Role with deleteBeforeReplace had an eks.AccessEntry dependency. The API calls executed by the aws provider looked like this:

2025-01-06T21:01:49Z           DeleteRole
2025-01-06T21:01:50Z           CreateRole
2025-01-06T21:01:51Z    CreateAccessEntry

The DeleteRole operation is eventually consistent, and in this instance it took several seconds to propagate through AWS. The CreateAccessEntry call thus succeeded with a reference to the deleted role, and the newly created role was unable to access the cluster. Since the ARN for the role did not change (we had changed the provider), the Pulumi state was wholly unaware of this.

GCP does [not have this problem]. I'm not sure about other cloud providers.

Related issues (optional)

• Closes [docs] deleteBeforeReplace affects dependents #13934