1st of all, sorry that i don’t have the answer for you. Some time a go I was doing something that could have used a similar effect, but i gave it up, since the restrictions and difficulties this effect would entail were IMO too great.
I don’t know how “seamless” we are talking about here, but bear in mind that because the different nature of the two lenses, a “seamless” transition would only be possible in very specific situation.
In a perspective view things appear smaller the farther away from the camera they are. They also “drift” towards the center of the screen the farther they move. In an orthographic view their size or position on the screen doesn’t depend on distance from the camera.
This would mean that for even a near seamless transition, all the objects in the view would have to be at an equal distance from the camera at all times.
The same problem presents itself with any non-flat object, because some parts of the object are farther from the camera than some others.
That’s why it’s impossible to make a perfectly seamless switch between the lenses if there’s, let’s say, a cube in the top left corner of the view. In orthographic view you would only be able to see the camera facing plane of it, in perspective view you would be able to see the down and right pointing sides too.