I could be mistaken but I think Web Feature Layer sends over data for each feature in the current extent, whereas a Map Image Layer send over tiled raster data representing the features in the current extent. In my experience Map Image Layers can be faster for data with a lot of features but they add extra complexity with trying to pull information from the features.
A click even on a Feature Layer pulls the data directly from the local memory. When you click on a Map Image Layer a request is sent to the server for information based on the coordinate where you click occurred.
If you provide more information about what you're trying to accomplish it would be helpful for advising you on which to choose.
Systems Analyst II
City of Ventura
