I am okay with this change. But it's a fixed cost that we will have to pay forever and it's not useful on the server side (AFAICT).

I'm wondering if there is an alternative, which is for your application where you need itempotency in POST requests, to subclass request, e.g.

class IdempotentRequest < Protocol::HTTP::Request
  def idempotent?
    true
  end
end

I don't have a strong opinion one way or another, but I'd like to hear your thoughts on this design, the relevance to the server side, and the long term cost of extending the default Request class here, vs the value to users.