Claude Code crashes with SIGSEGV after ~19hr uptime with recurring caffeinate subprocess activity

Environment

• Claude Code version: 2.1.205
• OS: macOS 26.3.1 (25D2128)
• CPU: ARM-64 (Mac16,8)
• Terminal: Terminal.app (parent process zsh)

What happened

Claude Code process ran continuously for ~19 hours (launched 2026-07-17 13:50:53, crashed 2026-07-18 08:49:14).

Separately, caffeinate -dims was run manually during this session (a routine habit for this user, used before without issue — unlikely to be the trigger on its own).

Unified system logs show a distinct pattern: dozens of short-lived caffeinate subprocesses (new PID each time) starting ~2.5 minutes after Claude Code launched and recurring every few minutes in bursts throughout the entire ~19-hour session, up until the crash. This doesn't match a single manual invocation and suggests something in the app's own operation was repeatedly spawning caffeinate.

Crash details

• Exception: EXC_BAD_ACCESS (SIGSEGV), KERN_INVALID_ADDRESS at 0x656c5261632f446f
• Termination: Segmentation fault: 11
• Faulting thread: 4
• Incident ID: DA062D91-634E-4BA0-A69E-1AD0CEC8D326

Impact

No data loss — work was committed to git prior to the crash.

Possible lead

Long-uptime session combined with sustained periodic caffeinate subprocess churn may point to a resource leak or state-corruption bug that surfaces after extended runtime.