fix(webapp): restore Postgres fallback for non-ClickHouse OTLP spans

Environments with taskEventStore = 'taskEvent' / 'taskEventPartitioned' (Postgres-backed runs not yet migrated to ClickHouse) were hitting a throw in buildEventRepository when the org-scoped ClickHouse factory received those store values. The throw happened inside the grouping loop of #exportEvents, causing the entire OTLP request to return HTTP 500 — which the OpenTelemetry collector treats as non-retryable, silently dropping the full batch.

Fix: guard the getEventRepositoryForOrganizationSync call in #exportEvents to only invoke the ClickHouse factory for clickhouse/clickhouse_v2 stores. All other store values (taskEvent, taskEventPartitioned, postgres) route directly to the existing Postgres eventRepository, mirroring the pattern already used by resolveEventRepositoryForStore and getEventRepositoryForStore in eventRepository/index.server.ts.

Also wrap the factory call in a try/catch that falls back to Postgres so any future unexpected store values in an OTLP batch degrade gracefully instead of failing the whole request.

Author: matt-aitken

apps/webapp/app/v3/otlpExporter.server.ts

@@ -24,6 +24,7 @@ import type { ClickhouseFactory } from "~/services/clickhouse/clickhouseFactory.
 import { clickhouseFactory } from "~/services/clickhouse/clickhouseFactoryInstance.server";
 
 import { generateSpanId } from "./eventRepository/common.server";
+import { eventRepository } from "./eventRepository/eventRepository.server";
 import type {
   CreatableEventKind,
   CreatableEventStatus,
@@ -120,10 +121,25 @@ class OTLPExporter {
         const routeKey = `${event.organizationId}\0${taskEventStore}`;
         let resolved = routeCache.get(routeKey);
         if (!resolved) {
-          resolved = this._clickhouseFactory.getEventRepositoryForOrganizationSync(
-            taskEventStore,
-            event.organizationId
-          );
+          // Non-ClickHouse stores (taskEvent / taskEventPartitioned) are Postgres-backed.
+          // The ClickHouse factory only handles clickhouse/clickhouse_v2 and throws otherwise.
+          if (taskEventStore !== "clickhouse" && taskEventStore !== "clickhouse_v2") {
+            resolved = { key: "postgres:default", repository: eventRepository };
+          } else {
+            try {
+              resolved = this._clickhouseFactory.getEventRepositoryForOrganizationSync(
+                taskEventStore,
+                event.organizationId
+              );
+            } catch (error) {
+              logger.error("[OTLPExporter] Failed to resolve ClickHouse event repository", {
+                taskEventStore,
+                organizationId: event.organizationId,
+                error: error instanceof Error ? error.message : error,
+              });
+              resolved = { key: "postgres:default", repository: eventRepository };
+            }
+          }
           routeCache.set(routeKey, resolved);
         }