Life sucks when you get an error which you have seen this error like plenty of times and when you start debugging it and run out of all the ways you have always followed to fix it and all those didn't fixed the problem.
With nothing left now to approach, wondering what may have gone wrong.
The fundamentals are shattered
...

We have an internal nuget package that wraps up the IConfiguration+ConfigurationBuilder for various .net core console/service apps (TL;DR, because people got creative), and it has a dictionary property for the common sections we use. AppSettings (for backward compatibility), ConnectionStrings, and ServiceEndpoints. If the need arises, I can add methods to return any type of object (no one has requested yet, we try to keep configs dead simple)

ex. var myDatabaseConnectionString = ConfigurationManager.ConnectionStrings["MyDatabase"];

Code review for someone who updated a .net framework app to .net core and they wrote their own IConfiguration wrapper for accessing the appsettings.json file, so I pointed out that we already had a library for that.
In the reply, he said he couldn't use our library because it had an 'AppSettings' property and since his appsettings.json file didn't have that section, he didn't want to cause a runtime exception.

OK, WTF...I even sent him a link to the documentation (includes explaining the backward compatibility part)...why the frack would you think because a property exists and you don't use it, that would cause some kind of runtime exception?
We have dozens of .net framework apps migrated to .net core with zero code changes and no one ever brought this up as a concern (because, why would they?)

Deep breath...ahhh...I respond that not having an AppSettings section in the appsettings.json file won't cause an exception, if you don't have one, don't need it, you don't have to use it.

He went ahead merged+committed his code anyway with his own IConfiguration+ConfigurationBuilder plumbing.

Code addiction is real kids...it's real.