Renderer crashes with TypeError: Failed to fetch when multiple background agents complete simultaneously #27440

Description

What happened?
When launching multiple background subagents (via task(run_in_background=true)) simultaneously, the OpenCode renderer process throws a TypeError: Failed to fetch and becomes unresponsive.

The error occurs in the message polling pipeline (fetchMessages → retry → request → fetch), suggesting the Tauri IPC bridge is overwhelmed when multiple agent sessions complete at roughly the same time and their messages flood the message queue concurrently.

This is not a network issue — the agents themselves respond successfully. The renderer simply fails to fetch the messages from the backend process due to the burst load.

Stack trace

TypeError: Failed to fetch
at fetch (oc://renderer/assets/main-Dyn5r1Jq.js:102672:45)
at request (oc://renderer/assets/main-Dyn5r1Jq.js:58972:24)
at async retry (oc://renderer/assets/main-Dyn5r1Jq.js:63561:14)
at async fetchMessages (oc://renderer/assets/main-Dyn5r1Jq.js:65380:24)
at async loadMessages (oc://renderer/assets/main-Dyn5r1Jq.js:65404:7)
at async Promise.all (index 1)
at async oc://renderer/assets/main-Dyn5r1Jq.js:65562:13

Observed behavior by concurrency level

Even 3 concurrent background agents triggers the issue, suggesting the renderer's message polling lacks backpressure or throttling when multiple sessions produce messages simultaneously.

Plugins

oh-my-opencode v3.14.0 (though the bug is in OpenCode's renderer, not the plugin)

OpenCode version

1.4.49 (Tauri desktop app, installed via auto-updater)

Steps to reproduce

1、Open OpenCode desktop (Tauri mode)
2、Run the /ohmy-opencode-agent-tester slash command (oh-my-opencode skill)
3、This skill launches 14 background agents simultaneously (6 subagent_type + 8 category) with task(run_in_background=true) and prompt="ping"
4、When the agents complete and return responses, the renderer throws TypeError: Failed to fetch and becomes unresponsive

The /ohmy-opencode-agent-tester skill document itself serves as a minimal reproducible case — it is designed to test all subagent connectivity by pinging them all in parallel.

Note: Even reducing to 3 concurrent agents still triggers the same crash, so the issue is not specific to 14 agents — any concurrent background agent completion burst overwhelms the renderer's fetchMessages pipeline.

SKILL.md

Screenshot and/or share link

No response

Operating System

Windows 11 Pro 64-bit, Build 10.0.26200

Terminal

Windows Terminal (OpenCode Tauri desktop app, not terminal mode)