v1.4.1

@harperfast/prerender-browser v1.4.1

Two worker-resilience fixes surfaced while load-testing the multi-process render-service
supervisor in Docker. No API changes for existing callers.

Fixes

• Survive render failures (the blocker). RenderWorker.run() logged render rejections with
  this.render(job).catch(logger.error) — a pino method passed unbound, so on any render
  failure the catch itself threw (Cannot read properties of undefined (reading 'Symbol(pino.msgPrefix)')), turning a logged failure into an unhandledRejection that killed
  the worker (a crash loop under the supervisor). Now wrapped:
  .catch((err) => logger.error({ err }, 'failed to render job')). Verified live in Docker — a
  browser-launch failure is now logged and the worker stays up.

• Graceful SIGTERM drain. Previously SIGTERM/SIGINT just setTimeout(exit, 1000), dropping
  every in-flight render (a job leaves Harper's queue at claim time and is only posted back at the
  end of render(), so each rolling deploy lost claimed-but-unfinished jobs until their TTL
  re-queued them). Now RenderWorker.shutdown() stops the queue consumer (abortable, so it doesn't
  wait out the idle poll), drains in-flight renders (Promise.allSettled, bounded by a deadline),
  then tears down. ErrorHandler drains via an optional onTerminate hook wired from
  startWorker, with a hard backstop before the supervisor's SIGKILL grace.

Notes

• No breaking changes; ErrorHandler options are optional. 20/20 tests pass.
• dist/ is built at release time (gitignored); the tarball workflow runs npm run build.

Consuming

"@harperfast/prerender-browser": "https://github.com/HarperFast/prerender-plugin/releases/download/v1.4.1/harperfast-prerender-browser-1.4.1.tgz"