The issue with the faulty units (the WiFi not being there) does not seem to be caused by a hardware issue. Having a shell on a “broken” 5.0.0.5 device, I managed to connect to my home WiFi with no issues. This does not help the UI and the “front” part of the system realize it has network connectivity, though.
Having a unit in a “locked” state can be salvaged from, temporarily, in a setup like mine, by wiping the sailfish/home logical volume and re-creating a plain EXT4 FS on it. That “resets” the deathloop back to the “first Wizard” state (then it goes to the “Second wizard”, then to “Spinner of death” again).
Inspecting the ext4 FS from a “Spinner of death” state, I saw an overly long log from fsck, resolving to an empty FS. Maybe the previously sounded theory about the second call of the Wizard corrupting the file system is close to truth.
Last edit: Looks like connman not being launched is the reason behind the UI giving up on connectivity. If connman is launched, the device is able to connect to everything.