>>This is weird UPDATE statement. I suggest to start from the SELECT with JOIN into a temp table and see how long that SELECT will work.
>>
>>If the select finishes in reasonable time, then still keep the temp table approach and just do UPDATE after the select based on PK.
>
>What would be the difference with this:
>
>

>UPDATE Client
> SET MyDate=
>  (SELECT COALESCE((SELECT TOP 1 Invoice.AddDate FROM Invoice
>  WHERE Invoice.NoClient=Client.Numero
>  ORDER BY Invoice.Numero DESC),'1899-12-30 00:00:00'))
>

>
>In that one, I also do a massive update where the inner select includes the relational of each update to be done (Client.Numero).
>
>This one works ok.

This is one to one relationship and in the other case it's not 1 to 1. But I would not write this UPDATE that same way also. I'd rather use CTE and ROW_NUMBER approach.

It's a bad idea to write UPDATE that way. Check out this blog post

Dear FROM clause...

This is one of my many favorites of his blogs.