Test what the change claims to improve.
Observe both successful requests and expected failures. A useful test demonstrates the contract: a correct response, an understandable error, or the preservation of important state.
If repeated experiments do not resolve uncertainty, stop expanding the test matrix and state the next single observation that would matter.
This section explains a specific part of the published reference. Test the behaviour in a controlled setting and keep the result separate from the original assumption. A clear explanation identifies the affected request or interaction, the expected outcome and the observation that would show a difference. Read the related pages to understand how this detail connects to the larger subject without implying an untested service or a commercial claim.