I think I found a few issues in SqlServerConverter.ReadGeometry, perhaps its because we're using different version of SQL Server, but I had to make these changes to get it to work:

In ReadGeometryPolygon I had to change
.Select(i => poly.STExteriorRing().STPointN(i))
to
.Select(i => poly.STExteriorRing().STPointN(i+1))
In the docs I could find STPointN starts with index of 1 (Not 0)

Similarly, in ReadGeometryMultiPolygon I had to add +1 to these lines because of the index starting at 1 and not 0.
var poly = mpoly.STGeometryN(j+1);
.Select(i => poly.STExteriorRing().STPointN(i+1))

Also, I do not understand what this line was doing:
rings.AddRange(rings);
It's just adding a empty ring. And outerRing is not getting added at all. I changed it to this, and it seems to work:
rings.Add(outerRing);

I seem to have found a logic issue with the SplitMultiPolygon method in GeometryConversions.Shared.Utilities.

The 1st issue that I found is that on line 19, where the ring is returned as points and IsCcw is being called. IsCcw executes as expected, but the if condition checks if the evaluation is true. In my opinion, this should be changed to
if (!IsCcw(ring.GetPoints()))
{
outerRings.Add(new Tuple<ReadOnlySegmentCollection, IList>(ring, new List()));
}
else
innerRings.Add(ring);
}

The 2nd identified issue, is that a System.IO exception is thrown at the end of the foreach loop that traverses the innerRings, on line 37

Hello, Morten- I have had the luck of having some really strange geometry in some testing I've been doing, and one of the polygons was reported by your code as having the ring orientation wrong. This was the result of the IsCcw() method using an incorrect index for the points used in determining the cross product. It computed an outer ring to be counter-clockwise when it actually wasn't.

Quick fix- change the index offset in line 72 in utilities.cs from 2 to 1.

Cheers!