While you're right that there are some practical issues with this sort of system, one should also recognize that the patent system doesn't current scale well either. You have to check the claims against every publication in existence. Since this isn't possible, this ultimately has to be fought out in court -- yet a granted patent still carries with it a presumption that the patent rights should be upheld.

One could, however, have a system where if two separate teams file patents before either are published, they should not have to license it.

Scientists do something similar, in that they have to log absolutely everything they do and publish the important parts. Biologists and psychologists are even more on the hook, since they have to demonstrate compliance with experimental ethics regulations. It's a bureaucratic hassle, but it works pretty well.

It actually doesn't work, since the only thing required for a patent is the ability for it to be executed. The patent applicant doesn't need to be the one doing it or need to be working with it. All the applicant needs is the knowledge that an expert in the industry believes it can be implemented.