Following on from #2710, which pointed out that previously the csharp lexer was not properly identifying new contextual keywords as keywords, I looked at some contexts in which the current solution might overgeneralise. Here is an example (not good C#, but an effective illustration):

As can be seen (and confirmed by viewing the output through the raw formatter), the current code is treating almost every instance of "file" as a keyword.

In case it isn't clear in the image, the code used to produce the image above is as follows:

file record X(int file);
file record Y(int blah);

public static class Program {
    static int file(Func<int,int> file) => file(13);
    static int blah(Func<int,int> blah) => blah(13);

    public static void Main(string[] args) {
        int file = Program.file(file => 42);
        int blah = Program.blah(blah => 12);
        X x = new(9);
        Console.WriteLine($"{file}{blah}{x}{args}");
    }
}

I wonder if the results could be improved by only recognising "file" as a keyword if it is followed by something like "record", "class", or "enum", and how many of the other contextual keywords would be susceptible to similar handling.