The limit of AtomSphere API (api.boomi.com) Rate of requests per second is FIVE requests per second and this limit is ACCOUNT specific.
If API usage exceeds the limit of 5 requests per second, API users receive an HTTP 503 response indicating that the server is temporarily unable to fulfill the request. This is documented in the user guide. Regardless where you run the process on, Atom or Cloud or Molecule, as long the as you try to send over 5 requests per second to api.boomi.com endpoint for the particular account, an HTTP 503 error will be received. Note this limit is NOT per node.
If you use an external tool or use HTTP Client Connector to make the API request and exceed the limit, you may want to add a retry logic in your process.
If you use the AtomSphere API connector or AtomSphere Partner API connector, the connector handles the HTTP 503 errors automatically by retrying a request up to 5 times.
You could also delay a process execution by adding a sleep time through groovy script. Refer to article 'How to delay a process execution using groovy' for implementation.
Note if you use AtomSphere API to trigger process executions, your process will NOT run when platform is unavailable, because api.boomi.com is essentially pointing to platform. As a best practice, it is always recommended to put the process on schedule. Scheduled processes will not be impacted when platform is unavailable.