AsyncOpTimeoutWithOpCtxDeadlineLater sets a timeout on the request but does not set a timeout code. As a result, the timeout code returned from NI::startCommand will context-dependent. In particular, if a timeout happens to be hit during connection acquisition (which is more likely on slower hosts like the sanitizer ones), a PooledConnectionAcquisitionExceededTimeLimit will be returned rather than a NetworkInterfaceExceededTimeLimit error, as expected.

To solve this, we should be sure to set a timeout code on the request so that it will be returned during any timeout, regardless of when it occurs.