ConcurrentScenarioRunner

Executes a list of Scenario instances concurrently and writes all results to a shared KSLDatabase.

Execution model

simulate operates in two phases:

Phase 1 — concurrent simulation. Each scenario is dispatched as an independent coroutine on SimulationDispatcher.default, a CPU-bounded dispatcher limited to Runtime.availableProcessors() threads. Every coroutine builds a fresh Model via scenario.modelBuilder, attaches an InMemorySnapshotCollector to the model's lifecycle emitters, and runs the simulation. All database-bound data is captured in memory during this phase; no database writes occur.

Phase 2 — sequential database commit. After all coroutines complete (awaitAll), the collected snapshots are written to kslDb one scenario at a time via SnapshotBatchWriter. Serialising the writes avoids concurrent SQLite access while keeping simulation time fully parallel.

Model isolation requirement

Every scenario submitted to this runner must use a ModelBuilderIfc that constructs a fully independent Model instance on each ModelBuilderIfc.build call. Scenarios built with the backward-compatible model: Model constructors of Scenario wrap a single shared instance and must not be used here — concurrent coroutines operating on the same model produce incorrect results and data races.

Error handling

If a simulation throws a RuntimeException, the failing scenario's coroutine catches the exception, logs it, and continues without affecting other concurrent scenarios. The failing scenario's Scenario.simulationRun will contain the error message via SimulationRun.runErrorMsg; its partial snapshots are not committed to the database.

Parameters