Minority unfiltered relay defeats filtering

Once a modest minority of nodes (on the order of 10%) relay a transaction class, filtering by the remaining majority is effectively defeated — the transactions percolate to miners regardless.

For (strongest version as argued): Transaction relay is a gossip network; a transaction needs only one connected path of permissive nodes to reach a miner that will include it. With v30-style permissive nodes past a small connectivity threshold, such paths exist essentially always, so majority filtering reduces to a symbolic gesture: miners receive the transactions, treat them as network-normal, and mine them unless they independently choose extra filtering work.

Against (strongest version as argued): Percolation to a miner is not equivalent to frictionless service. Propagation through a minority sub-network is slower and less reliable, degrades fee estimation and compact-block reconstruction for the filtered class, and leaves those transactions dependent on the continued existence and topology of permissive nodes and pools. Filtering also retains expressive and coordinating value: it documents operator preferences and keeps the door open to miner-level policy, which is where exclusion would actually bind.

Nuance: The percolation logic is standard network reasoning and directionally sound; the precise "10%" threshold is a heuristic, sensitive to topology, connection counts, and whether any sizable miners filter. The claim is strongest about eventual inclusion and weakest about equal quality of service.

Common misstatements: "Majority node policy decides what gets mined" (miners' policy decides; node relay only affects transport). "Filtering nodes partition the network" (nodes with different relay policies still agree fully on block validity and chain state).