The problem is pop science crap that tries to explain superposition and entanglement with complete word salad instead of "linear combination" and "product vector cannot be factored into tensor product of single-qbit states", respectively.
Anyway it is extremely doubtful that more programmers will need to learn how to program a quantum computer than, say, will need to learn how to program a GPU. They're both useful co-processors for very specific workloads. Programming will remain a fundamentally classical endeavor.
Taking the analogy of a GPU, how would you describe what a quantum coprocessor could do for a specialist (in, say, neuroscience or material science)? What kind of datasets would be appropriate? How should they think about what can be done with quantum processing? That's where I think there is room for metaphor development.
For materials scientists the value proposition is clear: quantum computers will enable efficient quantum chemistry simulations. Here is an overview paper: https://arxiv.org/abs/1808.10402
What are useful metaphors in the application or use of quantum computers?