When running requests, especially when running under high concurrency, the execution can run into failed requests due to the network. Even running on a single node (both client and server), we can run into port exhaustion.
Ideally we would like to handle those failures to not invalidate the runs. However this imposes a certain policy on the execution which might not agree with the server implementation.
When running requests, especially when running under high concurrency, the execution can run into failed requests due to the network. Even running on a single node (both client and server), we can run into port exhaustion.
Ideally we would like to handle those failures to not invalidate the runs. However this imposes a certain policy on the execution which might not agree with the server implementation.