Timeouts Are Policy Decisions Disguised as Configuration

A timeout value quietly encodes how long the platform will hold a lock, block a queue, or leave state ambiguous, which makes it policy rather than tuning.

Every timeout is a bet about how long ambiguity is acceptable. Most were set once during integration and never revisited after the workload changed.

What a timeout actually decides

• How long shared state stays claimed by a worker that may be gone.
• When downstream steps are allowed to assume failure and act on it.
• How much load a stuck dependency can transfer to everything else.

Use the architecture guide, integration guide, and examples to set timeouts that account for what stays ambiguous while the clock runs.