[Bug] Kernel panic on Intel Mac with AMD dGPU during heavy parallel Claude Code CLI usage - WindowServer watchdog timeout

Bug Description
Title: Intel MacBook Pro (AMD dGPU) kernel panics correlating with heavy/parallel Claude Code CLI usage — WindowServer watchdog timeout

Environment: MacBook Pro 16" 2019, Intel Core i9-9980HK, 32GB RAM, macOS 26.5.2 (25F84), AMD Radeon Pro 5500M + Intel UHD 630 (auto-switching), iTerm2, Claude Code CLI.

Observed: 2 full kernel panics within ~7 hours (2026-07-01 13:34:58 UTC, 2026-07-01 20:38:37 UTC), both panic(...): userspace watchdog timeout: no successful checkins from WindowServer (2 induced crashes) in 120 seconds. Followed by 20 AMD GPU Reset events over the next ~7.5 hours.

One panic occurred 55 seconds after invoking the Workflow tool under Fable 5 + /effort ultracode (multi-agent parallel orchestration mode). The other panic occurred during Opus 4.8/Sonnet 5 usage without Fable 5 active, so this is not exclusively tied to one model — but it suggests sustained heavy/parallel CLI tool-use load (multiple concurrent sessions, high-frequency terminal output, Workflow multi-agent spawning) stresses WindowServer/AMD GPU driver stability on Intel Macs with discrete GPUs, not observed on Apple Silicon.

Ask: Any guidance on reducing terminal-rendering / concurrent-process load from Claude Code on Intel + dGPU Macs, or whether Workflow's concurrency cap should be more conservative on such hardware.

Environment Info

• Platform: darwin
• Terminal: iTerm.app
• Version: 2.1.201
• Feedback ID: 3e45af88-e1d4-4781-a046-4feb9d5c75b1

Errors