Everyone's use case is different. 
For me the #1 reason I will be sticking with external logic processing is code portability. In the long run, 99% of my time is spent on my logic, and I do not like the idea that if my hub dies or if I decide to change to HA/z2m/new hub du jour I lose all of my investment in my logic.
Keeping it in node-red makes it portable between hubs. I've switched from hubitat to HA to z2m and back to Hubitat again since moving my logic to node-red. And each time within 1-2 hours I had 100% of my logic working again, and working EXACTLY like it did on the previous hub.
To me, that equals sweat equity protection and freedom from being locked in to the radio hub I chose to use. Can't do that with built in apps or RM.
The second reason I like it is that with node-red I get to chose the hardware it runs on. I want more capacity/faster processing? I upgrade my hardware. I want fault tolerance? I can run it in a docker swarm or high availability VM. Etc.
In fact, this weekend I moved my 3 node-red instances from my low powered docker swarm to my high powered ProxMox cluster. Only took a few minutes and I lost nothing in the process. Can't do that with built in apps or RM.
