fix(db): eliminate 'database is locked' plugin-load failures

Three-layer fix for the recurring 'failed to load plugin
@cortexkit/opencode-magic-context error="database is locked"' errors
that crash opencode startup when multiple processes cold-open context.db
concurrently.

Root cause: during a concurrent cold open (two opencode processes
booting, or a child session spawning while the parent is mid-checkpoint),
the WAL writer lock is held by a sibling. The first read in openDatabase()
— enforceSchemaFence() -> getPersistedSchemaVersion() — ran BEFORE
busy_timeout was installed (it was set later inside initializeDatabase()),
so SQLite returned SQLITE_BUSY immediately instead of waiting.

Layer 1 (storage-db.ts): Set PRAGMA busy_timeout=5000 immediately after
new Database(), before enforceSchemaFence()'s schema_migrations read. This
was the actual race window.

Layer 2 (dream-timer.ts): openTimerDatabaseOrNull() only handled
openDatabase()'s null return (schema-fence path) but NOT its throw path
(fatal open error). The throw propagated through
startDreamScheduleTimer() -> plugin() -> opencode's plugin loader, which
logged 'failed to load plugin' and disabled magic-context for the entire
session. Now wrapped in try/catch.

Layer 3 (index.ts): Defensive try/catch around the
startDreamScheduleTimer() await in the plugin entry, so even if a future
code path inside the timer setup throws, plugin() survives and the rest
of magic-context (hooks, tools, RPC) still initializes.

Author: OpenCode Agent

packages/plugin/src/features/magic-context/storage-db.ts

@@ -1394,6 +1394,14 @@ export function openDatabase(dbPathOrOptions?: string | OpenDatabaseOptions): Da
         mkdirSync(dbDir, { recursive: true });
 
         const db = new Database(dbPath);
+        // Install busy_timeout IMMEDIATELY after open, before any read. Without
+        // this, enforceSchemaFence()'s schema_migrations read below can throw
+        // SQLITE_BUSY when a sibling process holds the WAL writer lock during a
+        // concurrent cold open (two opencode processes booting at once, or a
+        // child session spawning while the parent is mid-checkpoint). The
+        // busy_timeout set later inside initializeDatabase() is too late for
+        // this first read. See the "database is locked" plugin-load failures.
+        db.exec("PRAGMA busy_timeout=5000");
         if (!enforceSchemaFence(db, dbPath, latestSupportedVersion)) {
             closeQuietly(db);
             return null;

packages/plugin/src/index.ts

@@ -209,7 +209,11 @@ const plugin: Plugin = async (ctx) => {
                 : undefined,
             ensureRegistered: ensureProjectRegisteredFromOpenCodeDirectory,
         };
-        stopDreamTimerRegistration = await startDreamScheduleTimer(timerRegistration);
+        try {
+            stopDreamTimerRegistration = await startDreamScheduleTimer(timerRegistration);
+        } catch (err) {
+            log(`[magic-context] dream schedule timer failed to start (non-fatal): ${err}`);
+        }
 
         // Start RPC server for TUI↔server communication (replaces SQLite plugin_messages bus).
         // `storageDir` is hoisted above so the auto-update checker can also use it.

packages/plugin/src/plugin/dream-timer.ts

@@ -59,7 +59,15 @@ let activeTimer: ReturnType<typeof setInterval> | null = null;
  * deep in embedding registration and throws a confusing TypeError.
  */
 function openTimerDatabaseOrNull(context: string): Database | null {
-    const db = openDatabase();
+    let db: Database | null;
+    try {
+        db = openDatabase();
+    } catch (error) {
+        log(
+            `[dreamer] storage open threw; skipping ${context}: ${error instanceof Error ? error.message : String(error)}`,
+        );
+        return null;
+    }
     if (!db) {
         log(
             `[dreamer] storage unavailable; skipping ${context} (the cache schema is newer than this binary supports — restart/upgrade OpenCode/Pi/Magic Context to recover)`,