well there is a difference if you expect one record since you're forced to believe all you're interested in is one row (IF EXISTS is a boolean condition) or if you actually go and check statistics first if you're going for a seek, scan or something else for all X rows of Y which get filtered, Joined and whatever else. Where your EXISTS ends is not where your Statement (execution plan) ends. So while the SQL Server might estimate for a direct call to the view that it should expect based on statistics for each ClovekAutoID to find 40.000 rows but if you tell me you are only interested if rows for a certain ClovekAutoID exist, then I would expect something different from you, too.
It's like telling "hey, I got a job for you" "Ok, great!" "Have a look if you have any book from XY" "OK, sure! - Yes, I do!" "Great, as you've done that so fast, go ahead and check from all possible books if you have the latest book from that Author" "OK, in this case as I was fast and I'll grab each book to check if this one is the latest one instead of checking through the whole shelf, eventually grouping all books from XY together and THEN decide which one is the newest."
I wouldn't even try to put this on either Cardinality Engine old or new, if you pass one parameter to check if anything exists, you should be looking for either literally anything or the last row inserted.
IF EXISTS (SELECT User_ID from SSCDB where FirstName = 'Jeff') is fine if you would like to pay any Jeff a beer, but knowing you might have to serve up to 40.000 beer because we're hoarding Jeff's might end up giving substantial, additional cost.