The checkpoint size buffer added in WT_DISAGG_CHECKPOINT_SIZE_BUFFER creates a two issues:

1. If by chance the database size were to go to zero, we'd re-add the buffer.
2. When the database size dips below the buffer that would technically be a bug but we won't necessarily catch it if it goes back above the buffer.

To fix this we should assert in __checkpoint_update_disagg_database_size that the database size doesn't go below the buffer size.

Scope:

• Modify the existing subtraction assert.
• Evergreen patch build