1. If one model becomes unresponsive, switch to an alternative. 2. Restrict the number of requests to stay below the limit. On the event of a failure, pause for one minute, and then retry.
If one model becomes unresponsive, switch to an alternative.
Restrict the number of requests to stay below the limit. On the event of a failure, pause for one minute, and then retry.