Do we have to see something to accept it with reasonable certainty as a fact - at least tentatively?
No. As per Ellis' example, in the course of our daily lives, although there are exceptions, we accept as fact (at least tentatively) most of the information we receive. We would probably learn very little if we felt compelled to question every item of information on every page of every book.
Do we have to be able to conduct reproducible experiments to consider something science?
If the experiment could be shown to be scientifically sound, then the results would amount to evidence. But if it couldn't be reproduced, then any theory based on its results would be in doubt. So the experiment is (good) science, irrespective of reproducibility, and the appropriate treatment of the results is also (good) science.
