zai-org / zai-org/feedback

Bug: Subagents silently time out (600000ms) after long parent sessions — spawn never completes

Open
#176 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

priority: P2
Dominant language
No language data
Stars
22
Forks
1
PR merge metrics
No merged PRs in 30d

Description

Reporter

Discord user: TiltStable
Discord thread: https://discord.com/channels/1346756824233148527/1529813063631376404


Symptom

Subagents (Agent tool with subagent_type) silently time out at 600000ms (10 min) after the parent session runs for a while. No error message — ZCode just returns "Subagent was inactive for 600000ms". The per-subagent exec directory (~/.zcode/cli/exec/sess_subagent_<uuid>/) stays empty (0 bytes) — ZCode never writes the subagent stdout/stderr logs, suggesting the spawn itself does not complete.


Repro Steps (reproduced 3× today)

  1. Start a long interactive session (~940 log lines / ~688 KB transcript, ~18 prior agent calls).
  2. Dispatch any subagent (fable-judge, code-reviewer — same behavior) with any prompt (even a focused 5-question verify prompt with all files prepared locally, NO WebFetch required).
  3. Result: "Subagent was inactive for 600000ms".

Worked Earlier in the Same ZCode Version

Same subagent (fable-judge) completed successfully earlier this week on a similar task: 335s, 49 tool uses, 718k tokens. Same ZCode version 3.4.2.3624 (installed 2026-07-21, app.asar modified 2026-07-21).

  • The same subagent (code-reviewer) ran today with a tightly-scoped prompt and completed in 57s.
  • With a broader prompt it hung at 600000ms.

What Was Ruled Out

  • WebFetch / network: Removed all WebFetch from the prompt, prepared all files locally with curlBash → local paths. Still hangs.
  • Process leak: Only 6 node.exe running, all are MCP servers (playwright, z_ai, context7 + npx wrappers), ~34 MB each, stable.
  • Memory: No OOM, plenty of free RAM.
  • File / network speed: curl on the same 87 KB file returns in 0.3s.

Hypothesis (low confidence)

Parent session context length / accumulated state affects subagent spawn. Correlation: longer parent session → higher hang rate. But this worked on similarly long sessions a few days ago, so something may have changed.


Environment

Detail Value
ZCode version 3.4.2.3624 (running; app.asar modified 2026-07-21)
MCP servers 4 configured (playwright, z_ai/mcp-server, context7, serena via uvx)

Regards,
Roman
http://www.rommark.dev

VIbe Coders Telegram Channel: https://t.me/VibeCodePrompterSystem

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the hang using a long parent session and the Agent tool, then inspect the per-subagent directory at ~/.zcode/cli/exec/sess_subagent_/ and the reported 600000ms timeout behavior. Compare a focused prompt that completes with a broader prompt that stalls; done means failed spawns return a useful error or complete normally with stdout/stderr logs.

Written by the indexing model from the issue text.

Assessment

Domain
ai-infra-agents, cli
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.