Great points!!  That's a great call out on the concat with NULLS unfortunately, I've had to resort to a case statement to set a default value '@@' for example.

I do agree having an ETL control framework to log success and failure steps is extremely important and saves a massive amount of time.  One thing that you might consider...

In your opinion, have you seen better performance with the data lake approach over dedicated SQL pools?

Why not generate a hash byte and compare the two tables based on the join condition?