A role sync with a safety net underneath it

One scenario fires when an employer is created and copies every role it has. A second fires on any change to a role. A third runs every fifteen minutes and looks for anything the first two missed. Each is small enough to read in a minute, and each can fail without taking the others with it.

The employer-created scenario sleeps for forty-five seconds before it reads anything, because the form writes the role subitems after the parent exists. It is a blunt answer to a race condition, and a cheap one: the run costs six operations, and anything that arrives later is caught by the other two scenarios.

The change scenario finds the matching role on the portal by name. If it exists, a router writes every mapped field onto it; if not, the same router creates it. Writing the whole set each time, rather than only the changed column, means a cleared field clears on the portal too. Changes to the column that records applications are ignored. It churns constantly and is not role content.

Every route checks that the parent employer's role sits in the open-jobs group before it does anything. Closed roles never reach the portal because nothing ever asks them to.

The sweep reads the forty newest roles and acts only on those created between two and twenty minutes ago. The two-minute floor lets the instant scenarios go first, so the sweep rarely races them. The twenty-minute ceiling against a fifteen-minute interval overlaps successive runs by five minutes, so a role cannot slip between two polls, and because every create is preceded by a name check, the overlap is harmless. A sweep with nothing to do costs two operations.