I don't understand the concern, at least if I'm reading the documentation for the speculative connect API correctly (first link in blurb).
All this seems to do is make the TCP connection (whether SSL or not) in anticipation of a link being clicked. The speculative connect API does not send any data in the TCP pipe it is creating. By opening the TCP link early, once the link is clicked, the TCP connection is probably ready to go, cutting down a bit on setup delay (which can sometimes be substantial if DNS is slow to resolve or the connection is using SSL), thus making the click seem more responsive to the user.
But nowhere in the docs is any mention of actual requests made to the server or any data downloaded from the server... until you click the link. Thus, the only information leaked by hovering over a link but not clicking on it is your externally-known IP address, which may show up in the error logs of the webserver as a dropped connection. There seems to be no danger of accidentally downloading a virus simply by hovering over a click.
If I'm missing something, please let me know.