In fact, instead of floating it off (which would involve massive problems with dimensional stability, frilling etc.), it's more likely in my mind that a transfer would be made to glass much like what we still do with carbon transfer. Gelatin has a good affinity to (very clean) glass. So the developed paper-film would be briefly soaked in water and squeegeed/stuck against a clean glass surface. This would then be brought into warm water, during which the image (hardened gelatin) would adhere to the glass, while the unhardened gelatin glue layer would dissolve, allowing the paper to float away. The image would be left on the glass and if done properly, this actually creates a very durable bond. The end result is essentially a glass negative very much like a dry plate glass negative.