Yes, restoring the database image is the bottleneck, a huge one. I hope too that some other approach will be created. Like continuing from the point in time aka from a specific block onward.

With the restore process itself, IOPS is the key. I don't think that parallelization has any impact. Disk speed is far more important.

With docker images and VMs you have additional layers between the process and the hardware that throttle it even further. I'll see how it will go with this setup above. And then when I move the whole VM to a NVMe disk :)

Have a great day.