Netgear's X-RAID2 is the same idea as Synology's SHR: drives of different sizes are sliced into equal pieces, each set of pieces becomes an mdadm array, and the arrays are stacked into one volume that grows as larger drives are added. ReadyNAS OS 6 puts Btrfs on top; the older RAIDiator units used ext4 and, before that, ext3. When the unit boots it assembles each layer in turn, and Dead is what it reports when one of them will not assemble: a member missing, a superblock unreadable, or an expansion that was interrupted half-way and left two geometries on the drives.
The drives, in every one of those cases, still hold the data and the metadata that describes the set. What the Remove inactive volumes button does is erase that metadata so that the unit can offer the drives as new, and that is the difference between a set the bench assembles in an afternoon and one it has to reconstruct by hand. Factory default does the same thing more thoroughly. Neither is a repair; both are the unit tidying up after what it thinks is a loss.
The bench images every drive, reads the mdadm superblocks from the images for each layer's order, chunk size and event counts, finds the reshape position where an expansion was interrupted, and assembles the layers in software. The Btrfs volume is then mounted read-only from the virtual set, with its checksums used to choose good metadata over bad, and the files are copied out. With Netgear gone, that is now the only route, and it is a good one.