An operator may receive several events about one entity. Retries can arrive while earlier work is still running. Two operators can also propose overlapping actions. Fibric places coordination controls in the kernel so each integration does not have to invent its own approach.
Concurrent attempts
Single-flight coordinates execution for an entity. It controls competing attempts during a defined execution window. The entity key must be specific enough to identify the work that should be coordinated, and the lock lifecycle must account for timeouts and failures.
Repeated requests
An idempotency key identifies a logical request across attempts. Reusing that key allows the execution path to recognize work it has already accepted or recorded. A new key represents a new request; changing keys on every retry defeats that purpose.
These controls have limits. A downstream service may accept an action and then fail to return a response. Work performed outside the governed path will not share its coordination. The connector still needs a way to reconcile uncertain outcomes, and the operator must not describe an uncertain attempt as completed work.
In a deployment review, define the entity key, idempotency scope, retention window, retry behavior, and reconciliation process together.