Woah! 2500w?? You trying to heat your house?
Seriously, you are gonna have a heck of a time trying to shed that kind of heat! Be prepared for a large setup and plenty of airflow. And I would do this kind of testing somewhere where an exploding lipo won't do any harm, because at those current levels, bad things can happen.
When I was in the design stage, I decided to stick to lower cell counts, up to 3s max, to keep heat down. Also keep in mind that if you design it for 100A @ 6s, then hooking a 3s pack will only draw 50A. Likewise, if a 3s pack will draw 100A on your load, then running 6s on that same load will draw 200A. So, you might need some way to create load "modules" so you can adjust the resistance as needed based on the cell count under test.
100A @ 6s will require a ~0.2 ohm load. At that current level, wire and contact resistance can make a big difference. If your wire/connectors introduce say, 0.05 ohms of resistance, that calculated 100A draw will be more like ~80A because of the added resistance.
One of the roadblocks I had was coming up with an accurate way to draw a specific current. Since I was testing to confirm battery discharge ratings, I wanted to be able to draw an exact figure. But since battery voltage will fluctuate over the course of the cycle and the amount of load placed on them, I would have to keep adjusting the ESC output to make sure the current stays constant. Easy enough if testing constant discharge, but when testing bursts, the fluctuating voltage will make it darn near impossible to manually adjust it. This would need a form of feedback to tweak the ESC as needed. It just got too complicated to be called "home brew". My point is: if you are doing this for your own benefit, then this issue is no big deal as long as you are "close enough" to satisfy you.