Ethereum scalability and Braess's paradox

With EIP 4844 integration seemingly closer and closer on the horizon it seems like we are about to witness the beginning of major breakthroughs in scalability, and this is only the beginning of the major changes to come. The ultimate goal of these upgrades is to reduce L1 gas fees via the empowerment of rollups, which will in turn reduce L2 gas fees. This is an obvious win given today's traffic, however isn't there a chance this will just cause a correlated increase in traffic, and bring fees back up to where they were in a once congested network? Similar to how adding more lanes on a highway tends to cause more traffic, who's to say it wouldn't happen here? 100,000 TPS is a lot, but that kind of traffic is a pretty average number for a major web service/application, and ideally Ethereum (even with rollups) is a place where many services of such scale can exist