The tested machine appears big and with 96 vcpu + 384 gb yielding 20k writes sounds too low. Jumping to 60k is very good. But still too small a throughput for what the machine is capable of.
Batching ensures you run at cpu & memory speeds and only pay significant latency for the flush - which usually linux kernel coalesces well if concurrent.
Yeah, with that machine spec, I think if anything, the post goes to show that LISTEN/NOTIFY does actually _not_ scale well. A CDC-based solution would give you a multiple of these numbers on much smaller (and cheaper) hardware.
Batching ensures you run at cpu & memory speeds and only pay significant latency for the flush - which usually linux kernel coalesces well if concurrent.