[BUG] A timeout or failure in one upstream MCP causes all other MCPs to stop working

[geneives233]

Description

When one upstream MCP times out or encounters an error, all other upstream MCPs also become unavailable.

The failure should be isolated to the affected MCP. Other healthy MCPs should continue to function normally.

Expected Behavior

• A timeout or error in one upstream MCP should not affect other upstream MCPs.
• Requests to healthy MCPs should continue to succeed.
• The failed MCP should report its own error without impacting the rest of the system.

Actual Behavior

• After one upstream MCP times out or fails, requests to all other upstream MCPs also fail or become unavailable.

Impact

This issue creates a single point of failure. A problem with any upstream MCP can disrupt access to every other configured MCP, reducing the overall reliability of the system.