Meta-level control knowledge sources — BB1-style stall detection and strategy switching

[mdproctor]

Overview

Implement BB1-style meta-level control: control knowledge sources that monitor whether problem-solving is proceeding as expected and can switch `WorkerSelectionStrategy` when stalled.

Part of epic #77 (Blackboard Architecture Evolution).

Background

BB1 (Hayes-Roth 1985) introduced a key insight: apply the blackboard architecture to its own control. Meta-level knowledge sources monitor the blackboard for signs of stall (no progress, circular activation, resource exhaustion) and intervene — switching strategy, adjusting priorities, or escalating to a human.

Current CaseHub: `WorkerSelectionStrategy` reasons about which PlanItem to activate but has no mechanism to detect stall and self-modify.

What to build

• A `MetaControlStrategy` interface (or extension to `WorkerSelectionStrategy`) that receives execution history and can:
  • Detect stall conditions (N cycles with no new CaseContext writes)
  • Switch to an alternative `WorkerSelectionStrategy` at runtime
  • Inject synthetic PlanItems (e.g. "request human review")
  • Adjust priority weights dynamically based on observed outcomes
• Integration with EventLog — stall detection reads recent `WORKER_EXECUTION_COMPLETED` entries
• A default `NoOpMetaControlStrategy` (current behaviour)

Research reference

BB1 meta-control paper; reactive scheduling with reason maintenance (ScienceDirect).
Connects to: #30 (epic), #77 (evolution epic)