Clause 8 makes the rule set manageable. It defines abstract starting points, builds a matrix of deletion classes, and shows how a cluster is allocated to a class and given a specific deletion rule. A deletion class matrix is a compression device, and like every compression it loses something. The point is to lose the right things. A class combines a standard period with an abstract start. If the organisation has six standard periods and five abstract starts, the matrix has thirty cells at most, and most will be empty. Every cluster then points to one cell. The specific rule adds the real event, such as the date the contract ended or the date the last invoice was paid. That two layer design is what makes the framework maintainable. A change to a standard period updates one row and every cluster in it, while a change to one cluster's event updates one rule. The matrix also exposes nonsense early. If a cluster lands in a cell with an abstract start that cannot apply to it, the allocation is wrong. A reviewer can test the matrix by picking a cell and asking which clusters sit in it and why. An empty matrix with a note that says every cluster is unique is a sign that the organisation has not done the compression, and the clause exists to prevent exactly that. The allocation record should name who decided, when, and which evidence was used, because the decision is a judgement about purpose and law, not a technical default.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.