observability-record

Create a scoped logger once with `logger.scope(LOG_CONTEXT)` using a module-level constant, then call it with a constant dot-separated name and a fields object, e.g. `this.#log.warn('cache.refresh.rejected', { url, attempt, status })`. Never interpolate values into the message string.

Choose the level by whether something broke: failure when the application did not do what it must, the level below when an automatic step did not fire, message for completed actions worth knowing, and detail for investigation-only data. A guest failing a validation check is below failure, not failure.

console.log has no level, no request number, and never enters the failure selection, so in production it becomes a line nobody will see. Use the scoped application logger instead so lines are grouped, scrubbed, and routed to the failure store.

Lines below the failure level reach the failure store only if their name is in the selected-names list, matched by name prefix. Add the line name to that list in the same commit as the code that writes the line, otherwise it stays only in container output.

Parse the caught error with a helper like `describeError(error)` so class, message, code, and trimmed stack are recorded in a consistent shape, and mask emails at the call site. Also set a wait limit such as `AbortSignal.timeout`, since without a timeout the catch may never fire.

Do not use it to edit the logger implementation itself or the request-context handling, which belong to the separate observability rule. It covers only adding new log lines in domain code, not changing the logging infrastructure.