Runaway renderer

[ntudor]

I hit what looked like a runaway renderer/process storm in Ego Lite on macOS.

Environment:

• macOS 26.5.2, build 25F84
• Ego Lite 0.4.6.13
• Installed from the current GitHub release path

Symptom:
Ego Lite became unresponsive and the host became sluggish enough that I had to restart. A ps auxww snapshot at the time showed many Ego Lite renderer processes all runnable and consuming CPU.

Observed process summary from the snapshot:

• 84 Ego Lite processes total
• ~76 renderer processes
• summed Ego Lite CPU around 950% according to ps
• summed resident memory around 8.1 GB
• top renderer: ~30% CPU, ~1.4 GB RSS, 604 minutes accumulated CPU time
• many renderer processes were in R state
• GPU process was also active at ~24% CPU with 477 minutes accumulated CPU time

Top examples:

• ego Helper (Renderer) PID 81612: 30.3% CPU, 1.39 GB RSS, 604:37 CPU time
• ego Helper (Renderer) PID 44699: 28.6% CPU, 694 MB RSS, 54:42 CPU time
• ego Helper --type=gpu-process PID 46673: 23.8% CPU, 68 MB RSS, 477:27 CPU time

This looked less like one heavy tab and more like a renderer/task-space leak or runaway task-space/session accumulation. I do not have a minimal repro yet. The incident happened after normal browser automation usage with task spaces/tabs over time, and the process count/CPU climbed until the machine became sluggish.

A useful mitigation would be:

• visibility into task-space/tab/process counts
• automatic cleanup of stale/inactive task spaces
• a configurable process/renderer cap or warning
• logging when task spaces spawn/close renderers