Name the service boundary
Your objectiveDistinguish a provider capability from your application’s outcome.
Understand the idea.
A managed service removes some operational tasks, but it does not automatically prove the behavior of your application. Identify what the provider supplies, what your configuration controls and what your code must still do. This is a decision model, not a universal allocation of responsibility for every vendor.
A worked example.
A service can run a task for longer, while the task still fails when an external dependency stops responding. The service capability and the application’s handling of that failure are separate acceptance conditions.
Try the reasoning.
For one service you are evaluating, draw three boxes: provider capability, your configuration and your application behavior. Put each known requirement in a box and mark unresolved boundaries.
Carry it into practice.
Write acceptance tests for the application outcome, not only the service setting.
Read the reference: IETF · HTTP semantics ↗