(1) I discussed this question recently with few colleagues from biomedical field, and, as far as I can see, major part of the field do not accept a non-reviewed experimental findings as a proof of discovery. This is mostly for two reasons: first, there is no single preprint server in biology. There are BioRxiv, F1000, PeerJ, ASAPbio, Arxiv q-bio, and probably more. And all these papers are not indexed in the main paper bases, like PubMed or WoS. So to keep up with the "pre-publications" one will need to check every single pre-print server. And this is too much. Second problem is the fear that if the field will start to acknowledge the priority based on the claim, without checking the rigor of experimental design, it may promote the sloppy data handling, because researchers will rush to put their flag everywhere. And we are talking about the very populated field, where many sub-fields have a clinical relevance or otherwise related to highly sensitive problems of human/animal health and well-being.
(2) Probably no, but this is the risk that one takes in research when he/she is going to conference, submitting grant application, and etc. The question is whether this risk provides some benefits in return.
(3) I still think that biomedical field, in general, would benefit from fast communications of the research finding behind the closed doors (i.e. without making scientific statements to the public). Even from the personal perspective, it is a way to faster position yourself as a researcher among other colleagues.