The current ODS operator has hardcoded replica counts e.g. rhods-dashboard has 5 replicas and modelmesh-controller & odh-model-controller have 3 each.
For small and single user instances of RHODS e.g. running on SNO this many replicas are not needed and impact the installation of RHODS as the CPU limits cannot be met unless the full production-grade core count is met.

This impacts the ability of users/customers to evaluate the platform. This has been mentioned multiple times on forum-openshift-ai so there is a need for this to be flexible.