Eventstore Performance issue


I have deployed Eventstore as a 3 node cluster on AWS, on r5a.2xlarge instances. We use the $by-category projection, and a persistent subscription on this projection. We see the following error when we try to write large amounts of data using the go sdk (around 600-2000 writes/sec)

“Leader [{leaderEndPoint}, {instanceId:B}] appears to be DEAD (Gossip send failed); wait fo
r TCP to decide.”

Node Config:

RunProjections: System
ClusterSize: 3
Insecure: true
HttpPort: 2113
IntTcpPort: 1112
ExtTcpPort: 1113
EnableExternalTcp: false
EnableAtomPubOverHTTP: true
DiscoverViaDns: false
ProjectionThreads: 4
GossipSeed: <instance2 ip>:2113,<instance3 ip>:2113
Db: /var/lib/eventstore
Index: /var/lib/eventstore/index
Log: /var/log/eventstore

Any help/advice would be greatly appreciated!

Sounds like a support question to me.