Wrong on the first point, right on the second. Institutional knowledge can't be easily regained. To build up the knowledge to, say, make a transistor, you need a bunch of people experimenting with a bunch of things. Published scientific papers and patents will get you part of the way there, but the final stretch is still up to you, including things like which equipment to buy, purity of supplies (and where to get them!), how long the chip needs to be bombarded by each kind of particles, how much air the cleanroom needs to move. All the tiny details. You have to discover them by trial and error. Actual chip manufacturing companies have found themselves unable to get good yield until they copied the floor plan of another working fabrication plant, and they still have no idea why that mattered, but that's an extreme case. Maybe nobody expected miniscule air contamination from one process step was affecting another nearby process step, and in the original plan they were farther apart.

Yes if you want to wire a neighborhood for internet you can skip DSL and go straight to fiber. That's not the problem. The problem is that nobody in your company knows how deep to put the fiber to minimize problems, how much redundancy is needed, how strong the mechanical armor around the fiber needs to be, how many fibers per cable to meet future capacity needs without excessive costs, which landlords are friendly to you, nobody has the right connections to city hall to get digging permits approved expediently, and so on.