Sending in 0.8 µs is scary fast, I can't even get my clock synced to that with a (crappy) pps, but that would be on a 10G board I guess which is probably not going to be available on a embedded system with resource constraints.
For the right combination of budget, application and team you can get a custom solution with much lower latency. Back in the days my team beat Solarflare hands down, but only for a single application. Don't remember the total number, but each additional 4 bytes of the payload added 5-7ns, IIRC.
Yes, of cause. If you have the big-budget and great team you can optimize your application very much.
But in Embox the user applications stay the same, and customization was very easy!