One argument against error handling in T-SQL is that there is already error handling in the calling application. To which I'll respond that there are unique error events within SQL Server best handled within SQL Server such as a deadlock (assuming we're talking about TRY/CATCH here). But not all processes within T-SQL have to have error handling. That said, a multi-step ETL process done within T-SQL, yeah, that's a place where I'd probably put error handling and validations that everything is working.