[BUG] Task tool subagent processes not terminated after parent session ends (Linux)

Description

Task tool subagent processes (claude --resume <session-id>) are not terminated when the parent Claude Code session ends (via crash, Ctrl+C, timeout, or normal exit). These orphaned processes accumulate over time and consume significant RAM.

Environment

• Claude Code version: Latest (via ~/.local/bin/claude)
• OS: Linux Mint 22 / Ubuntu 24.04 (kernel 6.14.0-37-generic)
• RAM: 16 GB
• Shell: Bash

Steps to Reproduce

1. Start a Claude Code session
2. Use the Task tool multiple times (e.g., subagent_type=Explore, subagent_type=Plan)
3. End the session (Ctrl+C, /exit, or let it crash/timeout)
4. Check running processes: ps aux | grep "claude.*--resume"

Expected Behavior

All subagent processes should be terminated when the parent session ends.

Actual Behavior

Subagent processes persist as orphans with this signature:

/home/user/.local/bin/claude --output-format stream-json --verbose --input-format stream-json --model claude-sonnet-4-5 --resume <SESSION-ID> --disallowedTools Bash,Read,Write,Edit,Grep,Glob,WebFetch,WebSearch,Task,NotebookEdit,AskUserQuestion,TodoWrite --setting-sources --permission-mode default

In my case, after ~1 hour of active use, I had 20+ orphaned subagent processes, each consuming ~400 MB RAM (total ~8 GB).

Evidence

$ ps aux --sort=-%mem | grep claude | head -20
# Output showed 20+ processes with --resume flag, ~400MB RSS each

$ free -h
# Before cleanup: 9.6 GB used, 1.3 GB swap
# After pkill -f "claude.*--resume": 3.4 GB used, 643 MB swap
# ~6 GB RAM freed

Workaround

# Kill all orphaned subagents manually
pkill -f "claude.*--resume"

Root Cause Hypothesis

The parent Claude Code process spawns subagents via spawn() but there's no cleanup mechanism when:

• Parent session crashes
• User exits with Ctrl+C
• Session times out
• Normal /exit command

Subagents should be tracked and terminated when parent exits (e.g., via process group, or explicit cleanup on SIGINT/SIGTERM).

Related Issues

• #15975 - Zombie node processes on macOS (similar but different process type)
• #7020 - Memory leak during subagent orchestration (memory during session, not after)
• #8382 - Parallel Task() memory issues (closed as duplicate)

Impact

• Memory exhaustion over extended use
• System instability (swap thrashing, OOM kills)
• Users unaware of accumulating processes