So you can definitely use this pre-T100. It works best with civs that can get an extra Great Scientist, namely Babylon and The Maya, but it would probably work with any civ.
I did a very rough test run, sub-optimal in many ways, using Babylon. #1 in science by t100, backed my way into the renaissance via steel, first to Printing Press, had Education Chivalry and Machinery by t102... I got 1 tech/turn starting on t90 or so until t108. My overflow got slightly bigger every turn until I hit the final tier techs (IE the last tier that the AI had researched) at which point it immediately started dropping. I think it peaked at 2600 excess beakers, and most of the techs at the last tier of medieval cost around 640, so I got 4 more techs before I ran out.
Just to make it clear how much extra beakers this earned me, I had roughly 70beakers/turn naturally, and my excess grew or broke even until I researched the 640beaker techs. So, with 70 beakers/turn, I was getting 320beaker techs every turn and still gaining roughly 100-200 beakers excess each turn. So, the amount of bonus beakers I got from the exploit was about 300-400. Sure it only lasted 15 turns, but that is just game-breaking, even that early. It's the equivalent of 3-4 extra GSes at that stage.
And, on top of that, I didn't even attempt to beeline. I did everything very suboptimally up to that point, and had very low production and growth, so I'm pretty sure if I had been playing better, and had stopped collecting the last tier of techs when my excess beakers stopped growing, then beelined Chemistry after Education, I would have had Cannons by t100 and Artillery by t130.
Basically, here's my take on early use of it. It's not as optimal as later use, because you run out of techs to get "for free" pretty fast when you overtake the AI. There's a critical mass where the overflow at least breaks even (roughly 8x the cost of a tech it seems) or will grow as long as the majority of the AIs have researched that tech. Once you pass them, you quickly run out of free techs because the overflow amount is small that early in the game.
So, if you can bulb 2 GS in one turn around t90 you can basically learn every tech the AIs know at 1 tech/turn + a few more. (I used babylon + liberty to get 2 GS and bulbed them the same turn I finished Compass)
Also this was in a game where I completed NC *after* bulbing my GS.
So, the potential, if you time it right, is huge. I think the optimal path is Pacal, so you don't sit there with a GS burning your gpt all the way from Writing. You don't want to avoid teching writing, so Babylon can't exploit this quite as much.
Here are some screens from t104 when my excess ran out. As you can see I was #1 in tech and almost had enough to complete all medieval techs. However, in future runs I won't bother to fill out the upper tier of Medieval. It's better to beeline with that excess tech.
Next time I'll beeline Education with Pacal, use the Liberty and Long Count GS the turn I complete Education, rack up all the cheap techs, and then see how far I can get towards Dynamite. I'm pretty sure I can get Fertilizer by t110. The trick then is to get enough population growth and production to feed your university slots and build cannons. Getting Artillery on t135 is pointless if you don't have cannons to upgrade.
Alternatively, it might be best with Babylon, but instead of beelining Education, hug the bottom of the tree and bulb your GS when you complete Metal-Casting or maybe Physics. Those techs are more likely to be unresearched, so you don't "waste" your research on techs you can get for free. More importantly, you've got less techs left to research between you and Dynamite when you get the overflow beakers.