next-server zombie process with 400-700% CPU, won't terminate

[epaynter]

Environment

• Next.js: 16.0.7
• Node: 22.20.0
• OS: macOS (Darwin 25.1.0, arm64)
• Package manager: pnpm 9.14.4

Symptoms

1. next-server process spikes to 400-700% CPU during
   development
2. Render times go from ~300ms to 50-100+ seconds
3. Process won't terminate even with pkill -9 -f next-server
4. .next directory cannot be deleted ("Directory not
   empty") while zombie runs
5. Only fix is to kill -9 <specific PID> the zombie
   process
6. Multiple zombie processes accumulate across dev
   sessions

Console output

Single item size exceeds maxSize
GET /page 200 in 52s (compile: 773ms, render: 52s)

Note: Compile times stay normal (~1s), but render
times balloon to 50-100+ seconds.

Process state during issue

$ ps aux | grep next
next-server (v16.0.7) - 416% CPU - 2.8GB RAM

Trigger conditions

- Large React client components (~2400 lines)
- Complex React context provider patterns (nested
contexts, useMemo/useCallback)
- HMR during navigation between pages
- Issue persists even after killing dev server and
restarting

Workaround

Using webpack completely resolves the issue:

next dev --webpack

With webpack: ~300ms render times, stable CPU, clean
process termination.

Additional notes

- Cannot provide reproduction repo (private codebase)
- Issue appeared in Next.js 16.0.7 with Turbopack as
default
- The Single item size exceeds maxSize warning
consistently appears before CPU spike