This ticket requires some research to understand whether different cost coefficients are required for clustered index scan or can we reuse those of collscan (and same question for a fetch on a clustered collection).