The one product I can talk about: Back in the day, Bingo Card Creator had one-click guest accounts. Their conversion rate was 2. Not two percent. Two. Ever. They were a cause of a stupendous portion of my support burden. (From the perspective of most of my users gradual engagement means "The Googles ate my work and now you have ruined the day of a room full of third graders, you monster.") The engineering to support them was fiddly, and ripping it out made the application better. (Despite several attempts to improve them I don't think I ever had near the UX work invested to make the experience not be awful. Again, gradual engagement UX is quite challenging. In particular, the handoff between guest accounts and "real" trial accounts is of paramount importance to my business but is meaningless to customers who have guest accounts until they get to school, at which point they will often discover, to their surprise, that failing to make the decision yesterday to give me their email address now means their cards are totally inaccessible. I never successfully figured out a way -- copy, design, workflow, etc -- to avoid having huge numbers of people fail at this use case.)
Discontinuing guest accounts increased signups of "real" accounts and also sales, if I remember correctly. You can eyeball the signup graph here http://www.bingocardcreator.com/stats/signups-per-day Apologies in advance for the unclear axes -- that page hasn't had the underlying code updated in years, and I didn't even consider "Hey if I run this business for forever eventually that axis is going to get crowded."