[BUG] `kagent-adk static` agent spins at 100% CPU on persistent HTTP MCP tool call failures

[rinormaloku]

📋 Prerequisites

• I have searched the existing issues to avoid creating a duplicate
• By submitting this issue, you agree to follow our Code of Conduct
• I am using the latest version of the software
• I have tried to clear cache/cookies or used incognito mode (if ui-related)
• I can consistently reproduce this issue

🚦 Impact/Severity

Blocker (this doesn't just render the agent useless, it as well impacts other processes in the same node)

🐛 Bug Description

Description
When an MCP tool call fails repeatedly due to a persistent connectivity issue (e.g. connection reset by peer), the kagent-adk static Python process enters a tight retry loop with no backoff or circuit breaker, consuming 100% CPU indefinitely.

Steps to reproduce

1. Deploy a kagent-adk-based agent with an MCP server tool configured
2. Make the MCP server unreachable at the network level (e.g. waypoint proxy that resets connections)
3. Send the agent a message that triggers the MCP tool call
4. Observe CPU usage of the kagent-adk static process

Observed behavior
The agent process spins at 100% CPU. It never surfaces an error to the caller or stops retrying. The only way to stop it is to scale the deployment to 0.

PID 1322590  95.1% CPU
/.kagent/.venv/bin/python3 /.kagent/.venv/bin/kagent-adk static --host 0.0.0.0 --port 8080

The underlying error on the MCP server side:

Post "http://get-weather.kagent:3000/mcp": read tcp ...: read: connection reset by peer

Expected behavior
When an MCP tool call fails persistently, the agent should:

• Apply exponential backoff between retries
• Enforce a maximum retry count or timeout
• Return an error response to the caller rather than looping indefinitely

Environment
kagent-adk: static mode
Transport: HTTP MCP
Platform: Kubernetes (kind)