You absolutely could build this with a “traditional” centralized service that eg. the GOBERMENT (or whoever trusted stakeholder) runs and operates, but then you have a single bottleneck that’s entirely dependent on a single stakeholder, and you still need to implement eg. audit trails for state mutations, access control, etc. etc.

So before anyone comes forward and claims that their innovative solution will improve on the status quo, I generally expect them to be able to describe the status quo. And here, you haven’t done so.

In the U.S., county recorders allow for anyone to record to the centralized ledger (and this is literally paper technology that long predates computers), and the transactions themselves are validated when necessary to resolve a dispute: one can only sell what they already own, and if they sell something they no longer own it. The law allows for certain types of involuntary transactions: foreclosures, execution, inheritance (where the owner can voluntarily prescribe some rules but doesn’t get to control the timing of when those rules get executed, and the failure to affirmatively write stuff into a will means that the inheritance falls back to defaults), divorce, partition, adverse possession, reverter, and then a bunch of special rules that apply to governments like tax foreclosures or eminent domain. And no matter what the actual papers say, ownership of land still must be enforced by a sheriff.

Which portions of this status quo should be decentralized? Or centralized? What would the benefit be?

In my opinion, real estate is the worst candidate for decentralizing the ledger.