But you're talking about chemical engineering of making glue, and luckily 3m (etc) have already done all that work.

All Tesla have to do is "buy the right glue", and then "make sure it gets put on correctly". That is a bit of production engineering but seriously it's not hard to lay down a bead of adhesive, place the glass in, and then check that it's been done.

I understand what you're saying but I disagree that it can be reduced to 'buy the right glue' and 'make sure it gets put on correctly' any more it does any more justice to a model CPU to say it's just 'take a transistor and replicate it 5 billion times on a hunk of silicon'. While that's technically true in both cases, it neglects to consider the level of technology that goes into all of the details to make it possible.

I'm not saying that attaching glass to metal is at the same level of complexity as semiconductor fab, but it's a far cry from using some Elmer's Glue to stick two pieces of construction paper together.

And yes, I'm pretty sure Tesla didn't create their own adhesive for the purpose, but I'm equally sure 3M doesn't have an off-the-shelf "Model Y Roof Adhesive, 500ml" product. Or perhaps they did, but the engineering constraints to develop it were those of normal passenger cars, sunroofs, etc., not taking into account the various forces acting upon a 4'x7' sheet of glass at highway speeds, or perhaps the temperatures or speeds of an industrial applicator robot (or conversely, the slowness of a human applicator).

To go back to your original assertion, no, it's not hard to lay down a bead of adhesive, place the glass in and check that it's been done. If in fact the failure mode was 'Failure to apply adhesive', well, that's a fairly egregious and easy to spot problem. If it's 'adhesive failed for unknown reasons after a short period of time', that's an altogether different problem.

Don't forget to add the nuances of automation, machine wear, etc.