Core diagnostics contract
Diagnostics and failure boundaries
Compare recoverable warnings, Tracker-managed failures, and synchronous public exceptions without conflating them with consumer callback failures.
1
Recoverable warning
A warning runtime event, a diagnostics sink call, and retained warning statistics are three independent observations.
Warning event
No warning event received.
Diagnostics sink for this experiment
Shows this run separately from the retained historical call count.
No warning experiment run.
Retained warning stats
No warning details compared.
2
Tracker-managed resolver failure
Tracker invokes the resolver, catches its exception, records rule-label-error, and keeps the marker with a null label.
Diagnostics output
No resolver error received.
Result
No resolver experiment run.
Retained error statistics
3
Invalid configuration
This is a synchronous public API exception. The rejected candidate does not replace the committed configuration or request a successful synchronization.
Thrown exception
No invalid configuration attempted.
Atomicity
No invalid configuration attempted.
4
Lifecycle error
A disposable secondary Tracker crosses its terminal boundary. The main diagnostic runtime remains active.
Thrown exception
No lifecycle error triggered.
Sequence
No lifecycle experiment run.
Tracked fixture and main runtime
Main Tracker is operational.