### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

There are four model services under my LiteLLM instance, but traffic to certain endpoints gradually drops to zero and eventually receives almost no further requests.

### Steps to Reproduce

def _get_available_deployments( self, healthy_deployments: list, all_deployments: dict, ): """ Helper to get deployments using least busy strategy """ for d in healthy_deployments: ## if healthy deployment not yet used if d["model_info"]["id"] not in all_deployments: all_deployments[d["model_info"]["id"]] = 0 # map deployment to id # pick least busy deployment, with random jitter on ties min_traffic = float("inf") for k, v in all_deployments.items(): if v < min_traffic: min_traffic = v # collect all deployments tied at the minimum min_deployment_ids = [k for k, v in all_deployments.items() if v == min_traffic] if min_deployment_ids: chosen_id = random.choice(min_de…