fix: add WAL checkpoint on test cleanup to prevent SQLITE_IOERR_SHMSIZE

When 88 integration test files run sequentially in a single-fork
Vitest process, the SQLite WAL file grows unbounded. Adding a
TRUNCATE checkpoint after each test app cleanup prevents the SHM
mapping from exceeding its size limit.

tests/integration/test-server.ts

@@ -184,9 +184,14 @@ export async function buildTestApp(): Promise<TestApp> {
 
   const cleanup = async () => {
     await app.close();
-    // Don't delete the temp DB directory here — other test files in the same
-    // vitest run share it.  The directory lives under /tmp with a random UUID
-    // and is cleaned up by the OS.
+    // Checkpoint WAL to prevent unbounded growth across sequential test files.
+    // Without this, the WAL/SHM files grow until SQLite hits SQLITE_IOERR_SHMSIZE.
+    try {
+      const { sqlite } = await import("../../apps/api/src/db/index.js");
+      sqlite.pragma("wal_checkpoint(TRUNCATE)");
+    } catch {
+      // best-effort
+    }
   };
 
   return { app, cleanup };

vitest.config.ts

@@ -23,8 +23,6 @@ export default defineConfig({
     globals: true,
     testTimeout: 30_000,
     hookTimeout: 30_000,
-    // Run all test files in a single forked process so integration tests share
-    // the same SQLite connection and avoid SQLITE_BUSY races on WAL setup.
     pool: "forks",
     poolOptions: {
       forks: {
@@ -44,8 +42,6 @@ export default defineConfig({
       ".worktrees/**",
       ".claude/**",
     ],
-    // These env vars are injected into process.env BEFORE test files are
-    // imported, ensuring apps/api/src/config.ts picks them up correctly.
     env: {
       AUTH_ENABLED: "true",
       DEFAULT_USERNAME: "admin",