Sure, it takes time to process all those vertices and render them down to the raster. And if you have enough vertices, they probably don't add any information to the picture at all. Like, if you have a river with 1000 vertices and map it into a 1000x1000 image, you might get one vertex per pixel, but probably you'll end up with lots of pixels with more than one vertex in them: a waste of vertices. And any vertices that formed a straight line when snapped into the image pixel grid would also be wasted: you can draw straight lines using just the end points.
Another bad thing that can happen with features that have lots of vertices is that, once you zoom in close enough that the vertices are drawing at a useful scale relative to your image, most of the feature is "off screen". So you have to retrieve the whole thing, just to draw a small portion of it. Think about retrieving all of Lake Michigan just to draw the Chicago shoreline.
And big objects can take a lot of network space! A 1K vertex object will have 16K bytes (plu s a bit of header space), and in hex will take up 32K. A 1M vertex object will take up 32 Megabytes! That's a fair amount of network bandwidth.
How big should objects be? Just big enough, and no bigger :)