Arbitration is a bottleneck typically. This could be a sign of success, or a sign that the host system is inefficient because it is generating too many disputes.

The problem is that any machine (blockchain, Ricardian contracts, your watch, my car) ... works until it doesn't. If we knew why it stopped working, we'd fix it. Therefore what is left over is the problems that the machine can't fix. And the people can't fix - because they would have bought a better machine if one was available.

Therefore we assume that there will always be errors. That require a specialist to fix. That's what arbitration institutionalises - a special path for determining how the fix is chosen. Implementation is a later step but we probably have that with BPs and the like.

I'm not sure if this addresses your question...