As a student, I make mistakes all the times, and stealing my idea will not face a heavy retaliation. I don't mind if someone come-up with concurrent works, but I do mind if my key idea is leaked. Many graduate students in my field therefore only use pre-print service if their papers are either publicly presented or conditionally accepted.
I am looking for a website that can do most of what Arxiv does, with an additional private option or limited visibility, for example an option to make the paper unindexed with search engines. This way, my paper cannot be seen for at least a few months. By doing so, I can at least claim that I am one of the first few researchers who got the idea right. And most importantly, I can privately correct my mistakes, if any. Publishing mistakes is embarrassing and sometimes harm one's career. Many mistakes will be revealed during the peer-review processes and I have a chance to correct them before publicizing.
Please kindly think twice before giving ethically and "politically correct" comments like no one will scoop me, everyone will cite me, and the scoopers will be punished. At least in my field, scooping and idea-stealing is commonly observable, and big-names doesn't care about publicizing pre-prints because they have resources to punish the scoopers. I do understand that keeping my work private might harm the community as the information becomes less transparent. However, it could be also possible that my un-reviewed flawed works misguide the community, and a paper reviewed will benefit the society more.
I personally wasted six months because of someone else's mistake in a published work.
I heard that, in molecular biology, there is a widely-used service to upload your research data like protein structure and sequencing to get a timestamp. The uploaded materials can be set to be private for a period, and fully disclosed later. A lot of people do use this function. This is not exactly what I am looking for but share some similar ideas.
If I remember correctly, there is an interesting paper studying the scooping patterns with the aide of this database.