The transaction pipeline
Intent to confirmation in seven stages, the contract each stage owes the next, and the specific way each one fails when the network is busy.
Open the noteOne swap is the unit of work an engine is judged on. Everything upstream is planning and everything downstream is accounting, but the seconds between a route quote and a confirmed signature are where cost, landing rate and correctness are all decided at once.
These notes walk that path stage by stage. Each stage gets the same treatment: what it receives, what it must produce, what it is allowed to change, and the way it fails when the network is busy rather than when your code is wrong.
What happens to a single swap between the moment a scheduler emits it and the moment a validator finalises it, stage by stage, with the failure mode of each stage.
Intent to confirmation in seven stages, the contract each stage owes the next, and the specific way each one fails when the network is busy.
Open the noteWhy the 1232-byte packet is the real constraint, how address lookup tables buy back account slots, and what a table costs you in setup, rent and coupling.
Open the noteCompute unit limits and unit prices as two separate levers, how the product becomes a fee, and how to set both from simulation instead of superstition.
Open the notePipeline problems look like architecture problems from a distance. A run with a poor landing rate gets blamed on the scheduler when the real cause is a transaction that grew past the packet limit and started dropping accounts, or a compute unit limit set once and never re-simulated. Isolating the stages makes the difference visible, because each stage has a signature you can inspect and a failure that names itself.