E2E testing - Figure out expected_output

Original Author: @dkirillov

We should figure out how the E2E testing behaves and the interface for it - specifically around `expected_output` and the various stage output cases.

Some things that come up are:
- How strict we are with tests.
- If the stage produces nothing, do tests _need_ to check for nothing.
- If a stage produces something, but we don't care, do we still need to check for it.
- If a stage produces something, but that something is dynamic between each run, how do we check for that (or do we).

Start by reviewing the E2E testing implementation and the expected_output interface. Define and document the expected behavior for empty output, ignored output, dynamic output, and strictness before making changes; done means the interface rules are agreed and represented by the relevant E2E tests.

Written by the indexing model from the issue text.