Description

When making multiple tool calls (specifically cron operations) during an active conversation turn, the gateway RPC calls timeout with:

[tools] cron failed: gateway timeout after 10000ms
Gateway target: ws://127.0.0.1:18789
Source: local loopback

Environment

• Version: v2026.1.30-38-ga5ef874e5
• OS: Linux 6.8.0-90-generic (x64)
• Node: v22.22.0

Configuration

{
  "maxConcurrent": 16,
  "subagentsMaxConcurrent": 32,
  "cronMaxConcurrent": 12
}

Reproduction

1. Start an active conversation with the agent
2. During the conversation, have the agent rapidly create/update/remove cron jobs
3. Example: Setting up download monitoring cron jobs while chatting

Observed Behavior

Multiple [tools] cron failed: gateway timeout after 10000ms errors in logs. The timeouts occur in bursts when:

• Creating cron jobs via cron action=add
• Removing cron jobs via cron action=remove
• These happen while the main session is actively processing user messages

Expected Behavior

With maxConcurrent: 16 and cronMaxConcurrent: 12, concurrent tool RPC calls should be handled without timeout, even during active conversation turns.

Analysis

Looking at src/gateway/call.ts, each tool RPC call creates a new WebSocket connection to the gateway. The hypothesis is that these RPC calls are being blocked by the main session processing rather than being handled concurrently.

The concurrency settings appear to apply to sessions, not to RPC calls within a sessions tool execution.

Sample Logs

2026-01-31T21:10:40.581Z [tools] cron failed: gateway timeout after 10000ms
2026-01-31T21:10:54.374Z [tools] cron failed: gateway timeout after 10000ms
2026-01-31T21:16:18.886Z [tools] cron failed: gateway timeout after 10000ms
2026-01-31T21:16:30.640Z [tools] cron failed: gateway timeout after 10000ms
2026-01-31T21:17:02.856Z [tools] cron failed: gateway timeout after 30000ms
2026-01-31T21:17:11.221Z [tools] cron failed: gateway timeout after 10000ms

All occurred during a ~7 minute window of active conversation with rapid cron operations.