You still have blocks (commits) pointing to other blocks (commits). Some blocks just points to multiple other blocks (think the uncle system in ethereum, which is just exactly that; temporary forks that are still included into the chain)
In the version you mean, how does the blockchain (git, or any other blockchain by your definition) respond when two valid blocks are canidates to be added to the end of the chain?
Is only one block the winner? How is this property maintained as the blockchain grows?
These are the questions and issues the blockchain others in this thread mean can deal with, while, for example, these questions don't exactly make sense for, for example, git.
Seriously though, it just seems like a question of semantics. Is a blockchain just a specific sort of Merkle tree -- thus making Bitcoin a blockchain plus a distributed consensus protocol? Or does calling something a blockchain necessarily imply some consensus protocol already.
It doesn't. But for all practical purposes at this point in time it might as well.
You can search right now and find programming 'blockchain tutorials', targeting the cryptocurrency community, that go on to describe only linked blocks chained together via some embedded hash of a subset of the previous blocks (meta)data. This is what others in various comments are calling a cryptograpically secured blockchain.
Other tutorials will set up a mempool (pending transactions) or get peers communicating. I've never seen peer discovery or any discussion on chain value (ie height vs accumulated difficulty) for chain selection.
We've always been able to decide how to add an element to the end of a (possibly cryptographically secure) chain of blocks.
What separates the two definitions of blockchain in this thread is exactly the distributed and decentralized choosing of the top block.