Cross-system audits

What usually creates the need for cross-system audits?

When the same fact is stored in k systems, there are k(k − 1)/2 pairwise consistency relations. With k = 5 that is 10. Nobody audits all of them. Everyone designates a system of record (the one whose answer wins) and audits the k − 1 = 4 relations to it: a star, not a mesh.

That is not a technical optimisation. It is the only structure in which a disagreement has a defined resolution. Without a designated authority, an audit produces a ticket that two teams argue about and neither closes.

A standing cross-system audit usually treats a symptom. Divergence between systems that hold the same fact is generated by dual writes: two independent writers, no shared commit. The audit measures the damage. The fix is a single writer, with the other systems derived from its log (for example by change-data-capture, which streams the writer's changes to the others). The organisations that run expensive nightly reconciliations are largely the organisations that cannot retire a dual write, and the reasons are political more often than technical. That is an honest statement about where the constraint really sits.

Compression: audits scale as a star around a designated authority, and mostly exist to pay for a dual write nobody can remove.