Guesstimated render times are always frustrating. The user sees the image rendered as a thin line that grows more rectangular from top to bottom. The total estimated time is calculated from how a particular render progressed so far. So if only the background had been rendered, that time will reflect how much time it would take if the entire image were background. The more materials the render starts incorporating, the more that estimated time is going up.
Theoretically, the user may render a smaller resolution and calculate how much time a larger image would take based on those results. There are enough variables that a novice like me will find the entire process maddening.
Oh well. Back to tangled meshes. Yes, my modeling technique needs serious work. After all, self taught means continuous self learning. Hopefully you can still tell that this object will be a mechanical griffin.