@lhci/cli autorun causes severe memory exhaustion and system crashes during SvelteKit SSR audits, even with --concurrency=1

When running lhci autorun (specifically npm run lighthouse:assert) against a SvelteKit application with Server-Side Rendered (SSR) routes, the process rapidly exhausts available system Memory/RAM and crashes the development environment (IDE/Node background processes).

Even after explicitly throttling the Chrome instances using --concurrency=1 and limiting the URL pool, the headless Chrome subprocesses spawned by Lighthouse consume anomalous amounts of system resources before timing out or forcing an OS-level kill.

Steps to Reproduce Initialize a SvelteKit 5 project with multiple SSR dynamic routes. Configure .lighthouserc.cjs to target ~20 authenticated routing paths. Utilize a puppeteerScript proxy to handle authentication state. Run npx @lhci/cli autorun --config=./.lighthouserc.cjs --concurrency=1 inside the IDE terminal (e.g. VSCode, Cursor) or standard bash. Observe task manager: node and headless Google Chrome for Testing instances stack indefinitely without garbage collection, leading to an Out-Of-Memory (OOM) crash. Expected Behavior When using the --concurrency=1 flag, lhci should strictly open a single Chrome tab, complete the audit, flush the memory/close the Chromium instance, and sequentially move to the next URL without bringing down the host OS or IDE terminal.

Current Behavior Chrome/Node processes stack and hold memory between route audits until the entire environment crashes, making local lighthouse:assert checks impossible on standard development hardware.

Environment Information: @lhci/cli version: 0.15.1 Framework: SvelteKit 2 / Svelte 5 Node Version: 22.19.0 OS: macOS Puppeteer version: 24.37.5