Phil Pirozhkov

Would you use `allow_any_instance_of(MyClass)` instead? Or do you test something different?

According to [the docs](https://relishapp.com/rspec/rspec-mocks/v/3-8/docs/working-with-legacy-code), `...any_instance_of` is generally not recommended and is considered a code smell. Also, I remember it was mentioned somewhere (code or docs) that it was causing trouble...

Fair enough 👍

@yujinakayama Fantastic job, thank you! > there will be a lot of massaging PRs and issues @JonRowe I have some time to spare in November/December.

Thanks for reporting! This is what I could find: - https://github.com/rspec/rspec-rails/issues/2478 - https://github.com/rspec/rspec-rails/issues/2451

So the `failure` is a `RSpec::Core::Notifications::ExampleNotification`. `failure_notifications` map through `failed_examples` with `ExampleNotification.for`. `new` is `private` in `ExampleNotification`. The only place where it's called is: ```ruby return new(example) unless execution_result.status ==...

Can you monkey patch `Bisect#bisect_formatter_klass_for` and return your own formatter class from it?

Eventually, if there's demand. I can't really remember this request popping up while I contribute to RSpec. It's a relatively small change code-wise, but that formatter part of the private...

@JonRowe Am I doing the right thing?

Green! I have seen `features/command_line/init.feature:25`, and `features/command_line/bisect.feature` (on JRuby) failing intermittently on CI. Might be flaky. Couldn't reproduce locally.