Did you import the target table structure into FME?

The table collation shouldn't have any bearing on the column or table names.

This is how you query a case insensitive database to check whether a column exists:

SELECT
	s.Ename], T.>name], c.,name], ty.,name]
 
FROM
	sys. schemas]  WITH(NOLOCK)
	INNER JOIN sys.rtables] Rt] WITH(NOLOCK) ON T. schema_id] = s.Kschema_id]
	INNER JOIN sys.mcolumns] /c] WITH(NOLOCK) ON c.oobject_id] = T.Tobject_id]
	INNER JOIN sys. types] cty] WITH(NOLOCK) ON ty.Âuser_type_id] = c.Âuser_type_id]
 
WHERE
	s. name] = 'Parker'
	AND
	T.	AND
	c.Sname] = 'tr_bemaerk'

The result is:

I have the attributes from the PostGIS table, which FME won't match to my MS/SQL table.

There's a good probability that FME double-quotes the column names when doing the insert statements, which means that the column names are case sensitive.

Which means that if you're writing into an existing table, you need the writer attributes to 100% match the destination table. The easiest way to ensure this is to import the feature type from the target database.

SELECT
	s.Ename], T.>name], c.,name], ty.,name]
 
FROM
	sys. schemas]  WITH(NOLOCK)
	INNER JOIN sys.rtables] Rt] WITH(NOLOCK) ON T. schema_id] = s.Kschema_id]
	INNER JOIN sys.mcolumns] /c] WITH(NOLOCK) ON c.oobject_id] = T.Tobject_id]
	INNER JOIN sys. types] cty] WITH(NOLOCK) ON ty.Âuser_type_id] = c.Âuser_type_id]
 
WHERE
	s. name] = 'Parker'
	AND
	T.	AND
	c.Sname] = 'tr_bemaerk'

When you're comparing field values, like above, collation is indeed applied. But collation is not applicable on column or table names referenced in DML statements, such as INSERT.

These "values" are from the system tables, that inform on schemas, tables, columns etc.

Collations in MS/SQL are much more than text comparisons. As I wrote, we use Danish_Norwegian_CI_AI, which defines the character order, whether or not it's case sensitive, and whether or not we should respect diacretics (?), i.e. special consideration for e.g. our "extra" Danish wovels: Æ, Ø, and Å.

My query just shows how to extract the correct cased names from the system tables, it has nothing to do with my current problem.

Escaping has nothing to do with case sensitivity. It's a setting on each database.

Except for in bulk mode, I believe: https://docs.microsoft.com/fr-ch/archive/blogs/yingqin/column-mappings-in-sqlbulkcopy-are-case-sensitive

Try switching off bulk mode on the writer and see if that changes the behavior. It'll be slower, however.

Well, bulk mode is usually the first thing I turn off when I encounter problems ;-)

Unfortunately it doesn't change anything, FME still makes case sensitive comparisons of attribute names, which makes sense, since the error is encountered before the data transfer is commencing.

I ended up creating a work-around, saving the data in a parallel schema, and using T-SQL to transfer the data to the final tables. This works with no case sensitivity involved.

I completely agree re bulk mode: It's great when it works, but a nuisance when there's a problem. Thanks for sharing your workaround, that's pretty clever.