I believe this is a realistically implementable thing that would make log search really efficient and light client friendly, without centralized services or any fancy math The basic idea is based on a solution that I implemented in Geth 7 years ago and is still being used to speed up log filtering (though the false positive rate makes it kind of ineffective now, an issue which would also be permanently solved by the structure proposed in this EIP).
The basic idea is based on a solution that I implemented in Geth 7 years ago and is still being used to speed up log filtering (though the false positive rate makes it kind of ineffective now, an issue which would also be permanently solved by the structure proposed in this EIP).