R# 2019.3 regression: argument validation

I annotate parameters which do not accept null with the [NotNull] attribute. However, since this does not prevent a caller from passing in a null value, I also validate the parameters. For example:

public void Foo([NotNull] Bar baz)
{
    if (baz is null) throw new ArgumentNullException(nameof(baz));
    ...

In 2019.2 and earlier, this worked perfectly.

I've just updated to 2019.3, and R# is now showing an "expression is always false" warning on every "is null" test.