$/GB is key
This sounds to me most like Kaminario k2.n with the addition of pmem and QLC in a denser box. I can see the similarities to Flashblade except this is entirely disaggregated and each flasblade node is basicallly an HCI node with compute and storage and it can only scale to 75 nodes vs 1000.
But THE MOST IMPORTANT thing imo is that they can do alllll of this for the same $/GB as spinning rust with how they handle the QLC. If so, HPC, big data, even backups have no reason not to use this.
And I have a sneaking suspicion that block access is coming. Either in new types of c-nodes or the same container.