I am surprised when people say that little output is required to get good at conversation. To my way of thinking, if I were to wait until I was so good at reading and listening that conversation only required a little bit practice, assuming that would be possible, then I would have done it wrong. Here are my reasons.
Listening is difficult. However, listening during conversation, a real native activity, is significantly easier than listening to other real native activities/materials. To not allow conversation at an earlier stage would be to restrict myself to less comprehensible material, or to non-native material. I find comprehensible real native activities/materials more interesting, so not conversing at early stages would be a minus for me.
My primary reason for learning languages is to converse with native speakers. It’s my main motivation. Forcing myself not to converse would dampen my motivation. You see a lot of articles these days implying that if you don’t have motivation, you can just manufacture it. That doesn’t work for me, at least not for something that is going to take thousands of hours. My motivation is strong, but it is finite, so I avoid doing things that dampen it, such as delaying conversation.
I chose option 1, lots of input and lots of output. You might be wondering why I didn’t choose 3. I’m not sure I could pull off 3. I feel that becoming good at reading, while much less time consuming than listening or acquiring the necessary vocabulary, is extremely convenient, if not required, for language learning. I can’t imagine trying to visualize words, study grammar, do flashcards, etc. without having passable reading. And many of my languages use non-roman scripts, which require much more time to get comfortable with ime. So while I might, in theory, get enough listening via a lot of conversation, avoiding reading would probably significantly stunt my learning.
One last thing regarding conversation. Although I start it early, I do not like to start it from day one. I could probably exchange greetings or something, but attempting a one hour conversation (standard length for me) with no previous knowledge would be frustrating to myself and my partner. I find that 2-3 months of self-study gives me enough of a base to jump into one hour conversations. I suck at first, but usually in a few weeks I feel that I have a “useable” level. And of course, the comprehensible listening practice that I get during conversation is a big boost to my learning from that point on.