Runtime Checks Testing

While expanding on unit tests for a project, I started to wonder if a
"pre-unit testing" system made any sense --basically a dynamic runtime
checks system. The idea is that for every class, you create a runtime
check subclass, in which methods provide real-time conditions. Then
there would be some mode switch that would automatically use the check
classes vs the regular ones. Eg.

  class MyClass
    def exclaim(y)
      y + "!"
    end
  end

  class MyClass::Check < MyClass
    def exclaim(y)
      result = super
      raise SystemCheckError unless result == "#{y}!"
      result
    end
  end

Now obviously this is a very stupid example, but it presents the basic
idea. This is sort of like double entry book keeping. We are creating
some real-time checks to ensure things are running along as we expect
them too.

What do others think of this sort of thing? How might a system like
this effect unit testing? Could it replace it? Or help reduce the
extent that unit tests are needed? Or simply compliment them? Or is
this a worthless notion all together?

One thing I notice, it has some upside in encouraging good coding by
encouraging small methods.

T.