When leader mode was first merged in WT-16033, we introduced a separate data validation task to increase correctness coverage. This task creates a layered table and a matching non layered table and compares them to validate correctness.

We should at some point move this logic into the normal format workflow. The proposal is to:

• Randomly enable mirrored tables inside format
• Automatically create a non layered table when mirrored tables are enabled
• Compare the layered and non layered tables during the test
• Remove the standalone format-stress-data-validation-test-disagg-leader task
• Expand or clone format-stress-test-disagg-leader to include this data validation work

This reduces test duplication, keeps validation closer to the main execution path, and simplifies our Evergreen file.
Scope

• Modify mirrored table randomisation logic to format (to 50% in disagg)
• Add non layered comparison table generation
• Replace format-stress-data-validation-test-disagg-leader wih format-stress-test-disagg-leader{}