I worked at the company building the inflight reel-to-reel tape decks for satellites and reconnaissance planes. The bearings were substantially smaller, around 1.5-2mm I.D.
Meeting requirements of high altitude meant special care preparing those for assembly. After the Branson ultrasonic-Freon cleaning tank, they 'air' dried in nitrogen, then precisely lubricated. That was done one bearing at a time, on the platen of an enclosed digital lab scale, with a pneumatic metering hypodermic, one tiny dot on each of 5-7 balls adjacent to the separator/ cage (takes a breath) under a microscopic viewer.
Neither bore or shaft press fit, a very slight push fit. Theory was, should the bearing, shaft or bore seize, it MIGHT continue rotating on one of the other diameters.
None ever failed; even the one recovered from Space Shuttle Challenger.