Slashing is a product decision

The protocol can slash a solver. The product has to decide when that is fair, and what it costs to be wrong in either direction.

Bonding is the easy half. You pick a number, the solver posts it, the contract can take it. Sizing that number is a product decision pretending to be a parameter. Too small and it is a fee. Too large and honest solvers do not show up, which looks like a dead marketplace and gets misdiagnosed as a discovery problem.

False positives are the expensive half. A fill that is late on an xStock is a failure; a fill that is late on a stablecoin transfer might be a RPC blip. If the slash does not distinguish those, you either under-slash (and the bond stops meaning anything) or you over-slash (and the honest set leaves). Settlement windows exist because I would rather encode the distinction than argue it.

A slash that cannot be argued in one direction will be argued in every direction.

Reputation as a price, not a verdict

This is the other reason reputation does not replace the bond. A score can lower the posted capital for a solver with a history. It cannot be the thing you confiscate. Scores are cheap to farm on a new address; capital is not. Let the score move the price of admission. Let the bond be what is actually at risk.

Corrections welcome. This note will be revised rather than deleted if I am wrong.