better handle case where leader can't evaluate transition

aggregator/src/aggregator/aggregation_job_driver.rs

@@ -462,21 +462,17 @@ impl AggregationJobDriver {
                 let (prep_state, message) = match transition.evaluate(vdaf.as_ref()) {
                     Ok((state, message)) => (state, message),
                     Err(error) => {
-                        // This shouldn't ever happen, because we'd never store a transition that
-                        // can't be evaluated. Most likely this indicates a programmer error (e.g.,
-                        // using the wrong VDAF) but we still record this failure against a single
-                        // report, rather than failing the entire request, to minimize impact if we
-                        // ever encounter this bug.
-                        info!(
-                            report_id = %report_aggregation.report_id(),
-                            error = ?error,
-                            "Stored transition cannot be evaluated",
+                        let prepare_error = handle_ping_pong_error(
+                            task.id(),
+                            Role::Leader,
+                            report_aggregation.report_id(),
+                            error,
+                            &self.aggregate_step_failure_counter,
+                        );
+                        report_aggregations_to_write.push(
+                            report_aggregation
+                                .with_state(ReportAggregationState::Failed(prepare_error)),
                         );
-                        self.aggregate_step_failure_counter
-                            .add(1, &[KeyValue::new("type", "invalid_ping_pong_transition")]);
-                        report_aggregations_to_write.push(report_aggregation.with_state(
-                            ReportAggregationState::Failed(PrepareError::VdafPrepError),
-                        ));
                         continue;
                     }
                 };