If a beforeEach hook fails, then all subsequent tests in a suite and all sub-suites are not run. For example:

// hook-test.js
describe("outer context", function() {
  beforeEach(function() {
    throw new Error("this is a failure in a before each hook");
  });
  it("reports the first assertion", function() {

  });
  it("does not report the existence of this test case", function() {

  });
  describe("inner context", function() {
    it("does not report report its existence in the output", function() {

    });
  });

});

reports only a single testcase, even though there are three defined:

$ mocha --reporter min hook-test.js
0 passing (6ms)
  1 failing

  1) outer context "before each" hook for "reports the first assertion":
     Error: this is a failure in a before each hook
      at Context.<anonymous> (test/mocha-test.js:3:11)

As outlined in #1043, this is the intended behavior for both beforeEach as well as other hooks. This makes sense from an efficiency prespective; after all, there is no point in actually running the testcases when it is assured that they are going to fail.

The problem with this is that when you're refactoring a large codebase, or doing something like upgrading an underlying library, or rehabilitating a codebase that has allowed its test suite to get out o sync.... work that might take days or weeks, this behavior alters how many total testcases are reported. So as you make changes, the reporting varies widely. We've seen the pass/total numbers of test cases jump from 95/220, to 4/8, then down to 0/1 in the case of a global hook failure, then back up to 35/160 in a matter of minutes.

This can be very disorienting, and it obscures your overall progress towards your goal which is a completely green suite where all tests are passing. The fact that a test is not run, does not mean that it doesn't exist, and that it is not important.

Rather than exclude a test completely from the output, it makes more sense to not run it, but still report it as a failure. That way, the shape of the test suite remains constant. If I have 225 testcases, then that's what I have, even if only 26 of them are passing. I at least know that I'm 12% there, which test suites are related by the same failure, and I can track total progress.

If this makes sense, it would be something I'd be happy to help implement.