2026-10-22

Trench lessons — 2026-10-22

Key takeaways
  • Jito/BAM (Oct 19): 30–40% of stake now runs private order-flow auctions; a new pattern backruns bursts of same-slot buys
  • Raydium CLMM: anyone can now open extra pools for the same pair and fee tier
  • Platform watch: nothing breaks

Jito/BAM (Oct 19): 30–40% of stake now runs private order-flow auctions; a new pattern backruns bursts of same-slot buysevidence: B official blog, Jito's own estimate

  • What it says: Jito estimates 30–40% of stake participates in some order-flow auction (OFA). These are effectively private mempools: the validator's sidecar holds inbound TPU transactions during its leader slots and lets partnered searchers trade around them. One reported pattern: bots backrun a sequence of buy transactions sent microseconds apart in the same slot, pushing price between the buys and selling after. Another OFA enabled "wide sandwiching" on several apps for weeks.

Raydium CLMM: anyone can now open extra pools for the same pair and fee tierevidence: scheduled Oct 22; B changelog via press, not yet verified live

  • What changed: Raydium's dev changelog lists an Oct 22 upgrade removing the admin permission needed to open more than one CLMM pool for the same pair and fee level. Raydium tells apps not to assume the main pool is the deepest, and not to trust a pool's owner field (it can be set to any wallet). The old create instruction becomes breaking for pool creators. Swaps are unchanged.

Platform watch: nothing breaks

  • Jupiter, Meteora, GMGN, Axiom, Photon, Birdeye: no execution-relevant change found for Oct 21–22. No new Token-2022 transfer-fee incidents found.

Sources

  1. https://bam.dev/blog/order-flow-auctions-the-new-private-mempools/
  2. https://solanahub.de/en/news/jito-direct-bundle-ingest-bam/
  3. https://satoshinakamoto.tv/article/raydium-clmm-pool-upgrade-22oct
  4. https://crypto.news/pump-fun-launches-pump-bridge-with-faster-crosschain-trades-and-fewer-errors/
  5. https://github.com/pump-fun/pump-public-docs/commits

Evidence grades: A = large dataset, B = official docs, C = vendor/thread/small sample. Research notes for education only. Not financial advice.