My own experience is with software that was free for a year before my first sale (though I made it available for sale pretty much near the start). So I did do what you are thinking. But I developed it entirely in public: the first version (released) took me 2 hours to write. This approach was great for countering my perfectionism.

(1) Yes. You don't need a free version to sell stuff. But what does "leveraged" mean in your case? If it's a network effect, where your offer has no value unless there are already users (e.g. a dating site), you would need some way to build up users.

(2) All products are incompletely tested... it's an opportunity to show customer service... and version 2. There's a question of degree here of course. You're right, you do have a chance to try it out on free users - although bugs are just as annoying for them as for mium/fee customers. Maybe the way to think about it is: are they paying for the benefit you offer, or for the absence of bugs? People complain, but they still buy - because they know no-one is perfect. Provided you treat them with respect, they may even like you more because of it.

(3) People don't like it if you take something away, especially if they started to rely on it. It creates bad feeling. I think a free trial (time limited) will give you the info you need.