This ends now. Even if I must drag the gods down with it. lets go over simple set of assumptions i am making @mert let me know which one doesn't make sense. I bill 10SOL per slot for consults. 1. the validator has to periodically release shreds [thus txns reaching the TPU have to pass through without delaying] - Network requirement/Social consensus 2. Backrunning any txn - Involves route searching across the chain state - takes 10-20 ms atleast [including latency even if co located] 3. A backrun can be done in two ways -> either the source is passed through and the searcher only gets a signal and backrun goes separately [CASE 1] or you bundle the source and backrun [CASE 2] 4. in CASE 1 [10-20 ms delay = txn will be several positions apart] CASE 2 [src,backrun will be n and n+1] 5. As multiple txns in this thread have shown, there are CASE 2s with this searcher on bald slots. 6. Thus to do Case 2; you have to hold the source for 10-20 ms 7. Which means, the price of execution [within users slippage tolerance] - would differ than what he could have gotten in case 1 8. CASE 1 is fine, CASE 2 is not 9. @toly and you both are trying to make a fool of your audience, letting them think you are doing CASE 1, but doing CASE 2. Because 2 makes much more money per backrun vs CASE 1 [coz your guy mriya, gets a priority lane behind the user] coming up with straight answers is rough, requires acceptance, balls, agency & courage. hope you'll use it this time. @trentdotsol @sonic_from_ny @toly @uriklarman @bencirc @0xrafal @alessandrod @niteshnath @bigtimetapin @helius
