A plan badly underestimates rows after a large data-distribution change. Which evidence should be reviewed?

1. Only source indentation
2. Only the TLS certificate subject
3. Relevant statistics and actual versus estimated cardinalities
4. Only the procedure name