Answer to question 1: We can train a YOLO network from scratch, but that would take a lot of work (and costly GPU hours). As engineers and data scientists, we want to leverage as many prebuilt libraries and machine learning models as we can, so we are going to use a pre-trained YOLO model to get our application into production faster and more cheaply.
Answer to question 2: Perhaps yes, but the latest DL4J release provides only YOLO v2. However, when I talked to their Gitter (see https://deeplearning4j.org/), they informed me that with some additional effort, you can make it work. I mean you can import YOLO v3 with Keras import. Unfortunately, I tried but could not make it workfullly.
Answer to question 3: You should be able to directly feed your own video. However, if it does not work, or throws any unwanted exception, then video properties such as frame rate...