I agree in theory; that many tabs are pretty difficult to manage. Unfortunately, a garbage collection cycle forces me to focus on garbage collecting (since I will usually have several important sites open among the trash), breaking my concentration on the research. It ends up being more productive to leave them open for a while, until they reach some threshold where I have to do too much work to find the important open sites among the garbage. That point for me is definitely more than 25 but also less than 100. I assume that it's different for different people.

On a typical day this means that in the morning I start with the session I left the evening before, say 25 tabs or so (usually project + whatever documentation I need for it), then as the day goes by the number of open tabs slowly increases. By the evening it probably is at its peak somewhere between 50 and 70 or so, then before I go to bed I catch up on my reading and close anything that doesn't have to do with work ready for the next cycle.