This is a question I've been thinking about as well. Although I don't have the insight of the original authors, here's my reasoning: Consensus module analysis is suitable for figuring out modules that would be shared / common across the independent datasets you're considering. For example, which modules are common between multiple celltypes of a tissue. Module preservation analysis, on the other hand, allows you to say if module X in one cell-type, how well preserved or disrupted it in the other cell type. This allows you to differentiate if a condition or treatment affects different celltypes in the same way or not.