Moin!
(You told me to comment in the forum - here i am. edit 
https://www.youtube.com/watch?v=Rr5HcBlesNQ)
I’d be interested in NVMe namespaces.
What weird an wonderful things could i do with NVMe NS’?
I haven’t had a hands-on kind of chance to play around with namespaces, yet. But i’ve read, that there’s “some kind of magic” in the controller/firmware, that’s equivalent to load balancing. You’d effectively get 2 PCIe/NVMe devices, like /dev/nvme1n0.. and /dev/nvme1n1.., and they’d behave like 2 separate devices, with all the redundancy speed boosts, like double the queues, and more iops. (i am aware, that it’d still be 1 single point of failure, 1 device)
I’ve always wanted to build a test rig with NVMe namespaces, instead of partitions. 
…i haven’t been able to allocate the budget, though 
edit2: i’m sure, fancy server platforms(HW) can boot those SSDs - but, what about consumer platforms? Let’s say, i have an SSD that’s capable of more than 1 namespace - could i dual-boot Linux and Windows off of different namespaces? Of course, i’d have to juggle the namespaces 1st - there’s a Linux CLI tool for that.
edit3: we might’ve had a little race condition amongst distributed worker threads here …there’s another thread on the same topic: Discussion of Solidigm's New D7-PS1010 and D7-PS1030 PCIe 5.0 SSDs