It isn't even that it's hard. I mean, it's been done before, many times. But there are processes here with a stack of key control parameters that can and will drift.

Panels will not be consistent - they never are (and you'd be surprised at the variation). Rivets will tumble in feed tubes or get trapped in escapements. Vision systems will get confused by ambient light or reflections at certain times of day. Weld spatter will accumulate on tips, and frequent tip dressing will change weld parameters (timing, pressure, current etc). Adhesives will be effected by temperature, humidity. Batches of sealer might not be consistent. Weld timers will fail. Power supplies will fail. Flow drill screws will deform, go off centre and be a bitch to get right. Laser welder lens will accumulate weld crud. People will mis-load parts. PIA bits will come in below the required quality. Tools will get shimmed wrong. Robots will be badly programmed.

All of this stuff will take time to get right. It isn't impossible, but it is very difficult to get a line performing reliably. Tesla will be integrating a lot of new technologies, and at volumes that they are not used to, in an area with no serious history of car building, in a high wage state. I doubt they can get the people to manage the line and maintain this equipment.