Starting with v0.5.26 of the Operator, I'm observing multiple vMCP pods get created/destroyed during the creation of a VirtuamMCPServer resource. It does eventually come up, but the pod cycling is unexpected. In v0.5.25 and prior, only a single pod is created, as expected.

A screen recording showing this is below. I don't see anything obvious in logs, other than repeated entries about the vMCP reconciliation.

Can be reproduced in the demo sandbox:

• Working as expected: main branch (has operator v0.5.25)
• Cycling issue: see renovate/toolhive-dependencies branch (PR here, the multiple cycling pods causes the validation script to fail)