Social consensus's suitability depends on the use-case. For example, social consensus is used to determine 'what is Bitcoin', and works pretty well for that use (the BCH hard fork notwithstanding).
For determining the state of the blockchain at some point in the past, it is probably effective, because the fact unambiguously revealed itself at some point in the past.
The trusted sources can even be Trusted Execution Environments that automatically output the state at t minus 3 months. There is very little room for factionalism, given deviating from the truth is so obvious when the whole interested world could see the objective truth three months prior and would have trusted sources they can rely on to relay that fact to them.
The problem is that with social consensus you can argue for the way that it should have been in addition to arguing about the way that it was. This is exactly what happened with the DAO. Socially, the Etherum community decided that the DAO attack should not have happened, so they made a new client that required everyone to upgrade, and then they altered history that had actually happened.
This of course resulted in two factions, one that disagreed with the change and one that agreed with the change.
If you are using Etherum and your client complains that there are two valid histories and then asks you to use social consensus to pick the 'true' history, do you feel confident in the platform? Especially if you know that you have thousands of millions of dollars that you wanted to put into the system.
Agreeing to use social consensus to enforce the protocol doesn't give a green light to using it to change the protocol. As the DAO and the BCH hard fork showed, having an objective proof of work consensus protocol doesn't protect the blockchain from a social consensus to change the protocol.
That being said, depending social consensus to determine what happened could make it more likely that a faction will use it to determine what should have happened. I personally don't think it will make it very likely, given the inertia of the original chain, but we'll see.
It didn't alter history. No transactions were reverted. It changed the rules by moving ETH from the attacker-owned contract address to a new contract address from which DAO investors could withdraw their funds. You might say that's semantics but in this technical context it's important, had the HF suggested altering history it would have gained less support IMO.
For determining the state of the blockchain at some point in the past, it is probably effective, because the fact unambiguously revealed itself at some point in the past.
The trusted sources can even be Trusted Execution Environments that automatically output the state at t minus 3 months. There is very little room for factionalism, given deviating from the truth is so obvious when the whole interested world could see the objective truth three months prior and would have trusted sources they can rely on to relay that fact to them.