The glass is made in two hemispheres and the contact surface is ground flat. There’s no glue or mechanical attachment; the hemispheres are carefully lined up by hand and then they pull vacuum. Atmospheric pressure alone keeps them mated prior to submersion. Then of course they are pushed together harder and harder as they descend.

The engineers talked a university into letting them use their hyperbaric chamber for testing. One of the prototypes failed at high pressure, which sounded like a bomb going off. The glass was reduced to tiny slivers, which the university was still finding in nooks and crannies years later. (The Nat Geo team was not allowed back after that.)

The probes were timed on the bottom by using thin steel wire to attach the weight. Basically the wire would rust until it broke, then the globe would float to the surface and start broadcasting a radio beacon for pickup. They adjusted the time by varying the thickness of wire.

They also stuffed the globes with LED lights and used them as “sun-like” sources of illumination for underwater cave photography.