That'd fail for three reasons.

First, scrapers would start running JS. Whether they're running chromium-in-a-box or something more clever doesn't matter. Compute is effectively free for them, and even shitty WebOS set-top boxes can probably run a stripped-down headless browser with a JavaScript engine.

Second, a full clone in browser memory is massive for something like the Linux kernel. Lots of browsers and pro-JS folks wouldn't have the resources to run that.

Third, assuming that scrapers are willing to play the JS game, suddenly you have a massively increased rate of full clones happening. Even if serving raw git is cheap for your backend, the bandwidth the elevated clone count drives is not.