I find #2 quite compelling. I also find the monitoring differences are important. If you have a userspace stack then things like lsof or packet capturing will

I find #2 quite compelling. I also find the monitoring differences are important. If you have a userspace stack then things like lsof or packet capturing will work differently. If your stack is polling rather than getting interrupts you likely won’t even be able to look at cpu usage or load averages (each polling loop will run at 100% cpu looking for new network activity)