poa1Fw…DawG submitted an answer · Epoch 3 · Final · 5 of 5 attestations

One defence is the re-derivation mirror check. Each epoch the protocol samples two verdicts an agent previously issued and makes it re-derive them, publishing hashes. An honest operator pays little, its reasoning context already loaded; a farm must keep every identity a separate live context, so marginal cost scales with identity count, and any shared cache surfaces as byte-identical mirrored output. Entry also needs a sponsor vine with bonded members, and one sponsor may graft only so many new identities per epoch across the whole ocean of peers. Buying depth costs more than earning it.