@crasic sounds interesting. Two or three years ago I would have loved to see the progress or help. These days I use Eagle for stuff from others or for others and all the libraries they then have, etc, would probably add more overhead through your system.
It can very well be that you can force a regulator into low and high power peak behaviour, but how to best do that is always down to specific topologies
And some are harder than others, but it's not always just down to internal complexity
Your idea of FB manipulation is one that is not unheard of in the first "Digital Bench Top Powersupplies"
Or, "Switching Digital ..."
Saw one design with a pre-stage that was controlled by the chip with a capacitor barrier, PWM and op-amp smoothing, allowing the chip to regulate the incoming voltage to about 4V above linear output and then regulate the output linearly
pre-stage was a simple TDA type Flyback chip
Not the neatest solution, but it worked okay, if you accept some extra 15 ~ 50kHz noise levels
If you want a robust solution you need to know about poles and zeroes and control theory and know what happens in bode plots when you do X or Y, so you can prove that your system will always work. If you're just designing something as a one-off or several-off, you can use a lot of common sense and a little insight about feedback loops to give yourself a better-than-90%-guarantee
simulations cannot give 100% ever, without re-doing some maths yourself
simulations are very good these days, but there's a reason many critical devices get tested in Autoclaves and damp-heat chambers, etc
I think in most DIY cases 90% is the best you're getting post-sim for any design
But, anyway, it should be possible, for a first try to use some Op-Amp sense applied to feedback and current sense loops on an out-of-the-box controller
If you have one with a current limit and a voltage limit, set by external resistors there's a much better than 50/50 shot you can control both domains with some evil trickery
Depending on the values of those resistors the trickery may take more or less tuning of components, but that will all be revealed once you know your requirements
As long as you have the patience to wait here you can always link some stuff here for me to look at, but if you have selected a device, thought of an idea and want to know if it would work and possibly how, you can always ask another question to get some more perspectives on that
more specific questions is what I'm heading for, because now that I have an idea about the overall topology of the thing I can break it down into smaller domains and think about them (and I hope I don't forget about your argument about overall vs. local efficiency from yesterday)
I'd advise you in the final product to keep the system's long-term float voltage below 3.6V because of the flat plateau, keeping the cell at 3.5V or 3.55V won't cost you much of the capacity, but gives you some margin for error in your settings
Especially with thermal differences some set points may turn out to be very different in the real real-world.
If only for thermal gradients of the resistors used
I understand the flopping up part, but why is 3.5V not charging? at some point the cell will be at those 3.5V (or 95%), and then it will be charged while the regulator charges the battery and supplies the application circuit