Really depends on you. I know some people who are all about fit. They'd be fine with publishing few or no papers, if they can do interesting-ish work in a nice environment. I met one who basically told me she no strong research interests at all, and just wanted a PI who was supportive and would let her graduate on time. To be fair, she not only had a BS and a MS from elite institutes, but was also pursuing a PhD at another elite institute and thus had some more postgrad options (i.e., finance/consulting companies that hire for prestige) than most of us.
Conversely, I know people who are all about hard accomplishments. They don't really care if they like their PI or group members, only that they are able to publish in great journals, and frequently. In your situation, they would be eternally tormented by "What if?" if they went with option 2. One of my mentors confessed that he benefited from PIs who were "slightly dicks", because they would push him and be brutal with him but not too much. If we go more extreme, I know someone who refused to date until he got a Science paper, and who basically said that nothing his PI did/said could be as bad as his parents.
And of course, there's how much you value scientific freedom. If you're someone who's like "Oh I need to do research on this subject", then you're better off not working with a micromanaging PI.
Personally, for me, it would depend on how dickish PI 1 is, and what his attrition rate is. Like there's a difference between "raising his voice and telling you your ideas are stupid but giving you suggestions" and "screams at you for thirty minutes straight until you cry and then encourages the rest of the group to dogpile on you". I'd be okay with the former, especially given his output, but not the latter. That's not a big price to pay for such high impact work. And on that note, if almost every single grad student or postdoc can graduate in a timely manner with that many great pubs, I'd just go for it. But if there's constant mastering out, people mysteriously quitting, then eh. It's all about your odds.