This reminds me of the difference between micro-benchmarks vs real programs. I would always include extracts from real texts of a few different types (novels, l

This reminds me of the difference between micro-benchmarks vs real programs. I would always include extracts from real texts of a few different types (novels, lists, technical writing, short paragraphs like in a playscript). But a test text like these are also helpful in exercising a few rare cases, like z, q, and certain ligatures.