The main query here looks a lot like the one that has been around for a while. I'm not entirely sure where it came from originally.
The only difference is...

Velveeta22 (6/2/2016)

---

cyp901 (6/2/2016)

---

Interesting article...

In general, my preference is not to...

Interesting article...

In general, my preference is not to have the database server connect to the API directly.

There are the obvious security reasons but there are also performance considerations....

ScottPletcher (5/31/2016)

---

As a work around, you can add dummy rows to the table until it occupies at least 8 pages of space, at which point SQL will switch from mixed...

I believe what Jeff Moden was trying to say is...

K. Brian Kelley (3/31/2016)

---

cyp901 (3/31/2016)

---

Depending on the size of active directory... This may not be the most efficient way to get the data, especially for group memberships. It has been...

Depending on the size of active directory... This may not be the most efficient way to get the data, especially for group memberships. It has been a minute, but I...

Would a local loopback linked server be an easier solution for this?

Bill Talada (1/26/2016)

---

Rarely if ever will query plans be available on a database given to a data warehouse or for importing. Otherwise running profiler or looking at DMVs might...

The comment about using query plans to determine the columns used in joins is definitely a worthwhile approach.

This article was great and I wish developers especially would think more about...

I'd love to start a centralized wall of shame website for vendors that deploy databases this way.

I have been using this frequently to import data to SQL Server from Powershell.

https://gallery.technet.microsoft.com/scriptcenter/4208a159-a52e-4b99-83d4-8048468d29dd

It works very well with only a few limitations. Definitely worth checking out.

Thanks for the clarification.

EXISTS is also safer than IN when you're dealing with a nullable column.

Have you checked the linked server collation settings? Perhaps you could also manually collate to the collation defined on the Oracle database.

It does seem like an issue related to collation...