Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

Heartbeat will run one time, then silently stop running.

Steps to reproduce

1. start v2026.5.18 with the following config:
   "agents": {
   "defaults": {
   "workspace": "/home/node/.openclaw/workspace",
   "models": {
   "zai/glm-5.1": {
   "alias": "GLM"
   },
   "zai/glm-4.7": {
   }
   },
   "model": {
   "primary": "zai/glm-5.1"
   },
   "heartbeat": {
   "every": "1h",
   "target": "none",
   "model": "zai/glm-4.7",
   "prompt": "Read HEARTBEAT.md (workspace context). Follow it strictly. Do not infer or repeat old tasks from prior chats. If nothing needs attention, reply HEARTBEAT_OK.",
   "isolatedSession": false
   }
   }
   },

The key part is "isolatedSession": false "target": "none". Wait until Heartbeat gets executed couple times. This will show durationMs low enough that makes it obvious that LLM wasn't called.
openclaw system heartbeat last
│
◇

OpenClaw 2026.5.18 (unknown) — I've seen your commit messages. We'll work on that together.

│
◇
{
"ts": 1779300801033,
"status": "skipped",
"reason": "target-none",
"preview": "⚠️ Agent couldn't generate a response. Note: some tool actions may have already been executed — please verify before retrying.",
"durationMs": 19,
"hasMedia": false
}
git diff --no-index -w sessions.json sessions.json_failed
...

• "estimatedCostUsd": 0.01847444,
• "endedAt": 1779275651268,
• "runtimeMs": 46130,
• "pendingFinalDelivery": true,
• "pendingFinalDeliveryText": "⚠️ Agent couldn't generate a response. Note: some tool actions may have already been executed — please verify before retrying.",
• "pendingFinalDeliveryCreatedAt": 1779275651467,
• "pendingFinalDeliveryLastAttemptAt": 1779286401028,
• "pendingFinalDeliveryAttemptCount": 3,
• "pendingFinalDeliveryLastError": null
  }

Configuration like that worked in April, though only with GLM, not with codex for example.

Expected behavior

Heartbeats should run, even if previous run got stuck, or should make it more obvious what should I verify that it will work.

Actual behavior

heartbeats after some time stops to call llm altogether

OpenClaw version

2026.5.18

Operating system

docker-compose.yml build (./scripts/docker/setup.sh)

Install method

docker

Model

zai/glm-5.1

Provider / routing chain

glm coding plan

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response