why engine api NewPayload needs to execute payload?

[cliff0412]

i am studying the op stack code base, i learnt that ForkchoiceUpdate engine api already build the block, why later on, sequencer node still needs to call NewPayload where the payload is executed twice. Since payload is proposed by sequencer node, why it has to be executed twice? will that slow down the execution layer? (even there is caching mechanism in EL)

[joohhnnn]

• ForkchoiceUpdated (with attributes) just kicks off building a candidate block. It’s a dry run, not an import.
• NewPayload is the only way to validate and import under the current head. You still need it.
• In the common case there isn’t a second full execution: the EL reuses its build cache, so NewPayload is mostly quick checks + commit.
• A re-exec only happens if the head changed, the cache missed, or the block came from elsewhere.
• Performance impact is small; cached imports are fast, and re-exec is rare.