Scheduled tasks keep Bun process alive after graceful shutdown

Environment

• Nitro: current main (eee0abd9fc4f729445f5b1a4da06fa78daf7a3da) and 3.0.260610-beta
• Preset: bun
• Bun: 1.3.14
• OS: macOS arm64

Reproduction

Configure one scheduled task:

export default defineNitroConfig({
  preset: "bun",
  experimental: {
    tasks: true,
  },
  scheduledTasks: {
    "*/5 * * * *": ["test"],
  },
});

Build and start the server, then send SIGTERM:

pnpm nitro build
bun .output/server/index.mjs &
pid=$!
sleep 1
kill -TERM "$pid"
sleep 10
kill -0 "$pid"

The HTTP server logs Server closed successfully., but the process remains alive and kill -0 succeeds. The behavior reproduces without Redis or other external connections.

Description

startScheduleRunner() creates a Croner instance with the default unref: false option and discards the instance:

new Cron(schedule.cron, callback);

After the HTTP listener closes, the referenced Croner timeout keeps the Bun event loop alive. Containers therefore require a forced kill during shutdown.

Passing { unref: true } to Croner allows the process to exit after the server closes without changing task scheduling while the server is running.

I will submit a focused PR with a regression test.

Start at startScheduleRunner() and inspect the Croner construction described in the issue. Verify the shutdown behavior with the provided Bun reproduction, then add the requested regression test and confirm that the process exits after the HTTP server closes while scheduled tasks still run normally.