How does the DGraph handle the transaction stack situation? And how does the DGraph balance the request?

When the transaction stack situation happend, what will DGraph do? Nothing?

In my cluster (6 alphas with 3 replicas), when the problem occurs, the server become ossified (or zombie? Doesn’t crash but can’t do anything. I don’t know which is the right description.)
Monitoring shows that only one alpha uses a lot of memory but has a low CPU utilization (about 15%) (call it AlphaA), the others’s memory and CPU usage are both low. Even the other two alphas witch are in the same group as AlphaA have no sign of balancing request. (I’m sorry I lost the screenshots. I can only describe them as best I can.)

Is it true that for a group only one alpha is processing the query/mutation request?

PS, I start the zeros and alphas directly on physical machines without docker or k8s.