TPB's magnet list in the end was about 40MB. If git were scaled to that sort of level, it would be far larger. Also there would need to be a lot of code written on top of git to automatically synchronize files between users. Git does not really solve many of the problems faced with a decentralized public database of magnet links.
A blockchain would be far worse. First of all, there would need to be some sort of currency backing it, and that currency would have to hold significant value to secure the network. A blockchain solves the problem of distributed consensus. A public distributed list of magnet links does not need such strong consistency guarantees as is needed for a financial transaction log.
Both git and the blockchain are useful tools but should not be used as a solution to every distributed computing problem.
The blockchain can solve a lot of distributed computing problems. It covers a very wide range of use cases, however, a large subset of those problems can also be solved with other algorithms which are far simpler, more efficient, and do not come with problems inherent to the block chain, the ever growing size, the wasted cpu-cycles, 51% attacks etc.
