It turns out that you had an "anomaly" at the end of your link.  Here's the correct link...

https://www.brentozar.com/pastetheplan/?id=B1Ons1TUn

no problem

here...

Heh... You DO know me.  That also means...

That's the article which had shaped my low expectations.  Good to know it's been improved.  It's also interesting Jonathan used GENERATE_SERIES in place of the non-itvf method of TOP...

According to some testing that I saw (IIRC)...

Crud... sorry.  I'd posted a link to the BPA for Windows, not SQL server.  I took the post down.

Looking into it further, I don't even know if such a thing...

Sounds to me like some connection zealot wrote a proc to drop all sleeping connections every night.

Do you understand how it works?

Peter "Peso" Larsson wrote a nasty fast Prime Numbers function years ago and I tested it to confirm it.  I'll see if I can find it in my...

AND?  What did you find out?

Thank you all for the valuable replies, but I think this discussion diverted from my original question: why my tally table's function had worse execution plan than my colleague's...

Peter "Peso" Larsson wrote a nasty fast Prime Numbers function years ago and I tested it to confirm it.  I'll see if I can find it in my archives this...

My question would be... why do you use a resource intensive, slow as a While loop recursive CTE as your sequence generator?  There are many better methods.  See the following...

We copied xmlrw.dll from a working machine to the /system32/   and /SysWOW64/ directories on our Azure SQL VM in order to get this working.  I think system32 was...

Thank you, Jonathan. Almost every step in the execution plan now marked as "No Join Predicate" and the whole function now costs 100% with my original tally one costing...