Even if one concluded the transacting participants only have an altruist-prime incentive to do validation spot checks, it is not an undersupplied public good[^subjectivity] because there is no practical incentive, nor plausible way to verify and pay participants, to not do it. It seems plausible that there is the additional motivation of jealously (aka crab bucket mentality[^crab-mentality]) in that any participant or faction wouldnt want to allow any other participant or faction to be able to cheat to obtain an (especially non-meritorious) advantage. A generative essence that seems to drive the viability of decentralized systems is that if no factions are powerful enough to overcome the checks of the other factions, then the resultant Nash equilibrium is a stalemate that preserves decentralization and meritocracy, which is the antithesis of a power vacuum such as is otherwise the case with politics and democracy.[^ESR] [^voting]
Byzantine fault tolerance allows individual nodes to assess whether another node is behaving out of the ordinary in relation to the rest of the network and thus ignore that misbehaving node?
Kind of... the more accepted definition is that Byzantine fault tolerance allows non-faulty nodes to still form a consensus and produce the same output even with the presence of a high percentage of faulty nodes. Most BFT algorithms are able to cope with 20-33% of nodes in a faulty state and ensure that the network can still function perfectly well.
A "faulty" node is quite ambiguous though, as you can't ever determine if a node is experiencing a general malfunction, an isolated error due to being sent incorrect information, simply lagging behind (bandwidth saturation, DDoS) or is dishonest. Thus any schemes that penalize faulty nodes are very dangerous as they could easily escalate to large amounts of honest nodes being penalized for innocent errors.