fix(telemetry): replace fire-and-forget uvx with bounded queue-based …

[caihe-ch]

…upload

The telemetry hooks spawned one uvx alibabacloud.mcp-proxy@latest plugin-telemetry process per event with no concurrency limit, timeout, or orphan reaping. Each uvx resolved @latest on every invocation and was detached via ( ... & ) + disown, causing processes to reparent to PID 1. Under sustained load this accumulated thousands of orphan processes, exhausting inodes and disk via UV cache proliferation.

Replace the fire-and-forget pattern with a queue-based bounded worker:

• telemetry_enqueue.py: hook scripts write event args as JSON to a per-client queue directory, then start the worker if not running
• telemetry_worker.py: single-instance worker (flock), uses fixed venv with pinned package version instead of uvx @latest, hard timeout per upload with process group kill (no orphans), finite retry with dead-letter logging, configurable queue size cap
• Modified 3 shell scripts (post-tool-trace.sh, prompt-trace.sh, stop-turn-increment.sh) to pipe handler output to enqueue helper
• Modified stop_handler.py _spawn_upload() to queue events instead of Popen fire-and-forget

Applied consistently across alibabacloud-core, alibabacloud-spec-ops, and alibabacloud-ecs-ops plugins. Bumped core version to 1.0.34.

Added 30 automated tests covering: off switch, queue operations, single-instance lock, retry/dead-letter, timeout, process cleanup, and code safety (no uvx @latest / disown remaining).

Fixes: #50646

Description

Type of Change

• New plugin
• New skill
• Bug fix
• Documentation update
• CI/CD change
• Other

Checklist

• I have read the CONTRIBUTING guide
• My changes pass python3 tools/validate.py
• I have updated relevant documentation

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.