Invariance & Causal Representation Learning: Prospects and Limitations