If you're concerned about decentralization, SPV Fraud Proofs are better than ever: https://bitcrust.org/blog-fraud-proofs
I had never heard of this fraud proof implementation. Interesting.
I just started reading it and one problem I see is that it requires a fork to Bitcoin that changes the merkle tree. This would break covert ASICBOOST (I'm not joking!) so I don't think Jihan will support it.
It seems like to do what they call 'fraud hints' would require a malleability fix too, so we'd still need segwit or another malleability fix, though I'm not entirely sure about this.
There are also some limitations to it:
The tricky frauds to proof, are not covered this way. An attacker can include a TXID of a non-existent transaction in the block. Such absence cannot be proofed as there is no invalid transaction to broadcast.
This article does not aim to structurally cover the full attack surface, nor does it provide detailed solutions; it should serve as an initial exploration of the possibilities of Fraud Proof SPV nodes using the spend tree.