Should be tracking metrics such as, but not limited to

• Replication lag
• average wtime
• Average write latency
• Write throughput
• Cache fill ratio
• Other WT eviction statistics
• Secondary read latency
• How often is eviction falling behind
• CPU usage
• Active writer threads
• Apply and write buffer sizes

to determine if changing replWriterThreadCount has a positive / negative / neutral impact on the fleet (only for clusters > M60)