> Look at some civil engineering disasters: licensing doesn't prevent disasters.

I'd suggest that you are judging the effectiveness on the wrong criteria here. It's impossible to foresee all distasters - the goal is to enact mechanisms to mitigate the human cost of distasters, and even more importantly, prevent the second and third occurrences of the same distaster... (which we seem decently good at)

Licensing doesn't prevent similar disasters from happening. Licensing is simply too low-level to prevent problems. Licensing is a mechanism to scapegoat individual engineers. It doesn't fix issues with systems. And the penalty of losing a license is simply too cheap compared against the cost of major accidents.

Engineering projects are usually performed by businesses, often across jurisdictional boundaries. Licensing is usually a monopoly within a jurisdiction.

Avoiding accidents requires solutions at the level of a business activity: and it needs to cover all the aspects of the activity (not just the coalface of engineering, but the management and sales and every other aspect that can affect safety).

If software engineering requires licenses then open source cannot be used. Good luck starting from scratch and introducing other risks of proprietary solutions!

An example: a Lambourghini tractor has a fault in the steering software that causes the tractor to overturn and kills a farmer in Nabraska. How do you expect software engineering licensing to prevent that?