ara/ludicrous-mode-backend

[anthonyra]

Adds the backend magic for ludicrous signing mode.

For ludicrous mode (transport RPC) - it utilizes a custom rebroadcast logic for txns. Currently the normal method submits the txn and waits for the response from the waitForTxConfirmation which has been known to fail due to the blockhash expiring on block height. This happens after ~150 blocks (1min - 1min 21secs or even longer depending on congestion).. now for ludicrous mode we set { skipPreflight: false, maxRetries: 0, preflightCommitment: 'confirmed' } on sendTransaction and we use a custom signatureConfirmation based on the same preflightCommitment. If this signatureConfirmation fails we gracefully catch the ERROR and simply try and send the txn as long as the blockhash hasn't expired. We could incorporate the ability to set a list of RPC endpoints that it can round robin on each rebroadcast attempt as well. If we fail the sendTransaction due to the preflight checks or blockhash expiration we simply rebuild then sign and repeat. This is the quickest while being the safest approach I've seen based on the Solana documents about how transactions are handled by the blockchain.

Pros:

• ensures we don't have our txn processed on a fork
• 12 second (vs +1min) delay in knowing if txn was processed
• allows us to attempt to send the txn to different endpoints
• we can catch errors and gracefully handle the responses
• profits

Cons:

• more code
• could cost user's if setting multiple fallback RPC's

References

https://docs.solana.com/integrations/retrying-transactions
https://solanacookbook.com/guides/retrying-transactions.html#customizing-rebroadcast-logic

Where the 150 blocks expiration comes from;
https://github.com/solana-labs/solana/blob/53c723ae3f6b3013b9bfe275ea683ef2088737b3/sdk/program/src/clock.rs#L129