Why are you and others so against encoders?

Because your machine should be 100% reliable if set up correctly. Adding encoders won't make it any more reliable, they'll just give you the opportunity to stop it when things go wrong.

If you have a NEMA 34 or larger motor you have literally 5 times the torque I do. 380 oz-in motors don't have much overhead when driving against a 1/2" bit mounted in a router plowing through hardwood plywood. Feed rate changes as little as a few IPS under those conditions make the difference between a toasted bit and a successfully cut part.

The issue here is that you have a machine capable of a certain level of performance. You want it to perform like a machine with much higher capabilities. Encoders aren't going to do that. The "overhead" that you mention is up to you. The reality is that if the machine stalls or loses steps, then you should be running at least 15%-20% slower than you're trying to. That's how you get a reliable machine.

As far as burning bits. Try slowing down the rpm. As far as achieving recommended chip loads. You'll need at least a 5-7HP spindle to achieve those, as 3/8 spiral router bits are easily capable of cutting 3/4" deep in one pass at 800-1000ipm. With a normal router for a spindle, you'll stall it instantly at those feedrates. When you have an underpowered machine (which most of us have), you run at reduced depth of cut and lower rpms to keep the bits from burning, while getting the most you can from your machine and it's capabilities.

Brains will not help you do what you want to do.  Using encoders you can only detect gross position errors, unless you add custom hardware to monitor BOTH the encoders and the step pulses coming out of the parallel port, and determine for yourself whether you are really on-position or not in real-time.  This question comes up on a regular basis, and what you propose has NEVER succesfullly solved the problem.  In the end, everyone either gives up in frustration (after having spent a HUGE amount of time, money and effort in the attempt), or they finally break down and FIX the machine hardware.  If you have NEMA23 motors, you can buy higher torque motors for about $50 each.  More likely, the problem is your stepper drivers, and power supplies, which can be swapped for better ones for probably about $300 for all three axes.  SOLVING the problem will give you a machine that performs FAR better than one that simply detects the problems, and in the end you'll be reliably turning out MORE good parts MUCH sooner.  You got people here with many, many, many years of experience with CNC machines, and Mach3.  Ignore them at your peril.

Regards,
Ray L.

I have closed loop (encoder feedback) on 3 of 3 machines and feel uncomfortable without it.  I use it as an alarm to tell me something is wrong with the machine.  When closed loop is telling me that things aren't right, I investigate and resolve the issue.  Encoder feedback runs through a plug-in and ESTOPs when the difference between commanded position and encoder position exceeds the tolerance.

Your machine is not capabale of doing what you want it to do. So quit pushing it, modify the way you work, or
upgrade it or even get something different. If you have backlash but want more accuracy or assurance of position
when  machining then get rid of the backlash. You don't need encoders to find the backlash. If the backlash varies
then you have additional problems which only axis component replacement can cure. You have what you have and
if dosen't work for you encoders will not be of any help.

I am not against encoders. I have a digital display unit on the mill for each axis, so similar to having encoder readout, but, provides
nothing of real value other than confirmation that i am at some point / home or whatever, BUT, the Mach DRO is more accurate
than the digital readout and just as reliable becuase i have confidence in my machine and the way it is run.
Do as you wish, I don't think you'll learn any faster.

I don't have big steppers on any of my machines. I replaced the old 34's with 23's. I do have the axis adjusted such that I get
 max torque out of them while still minimizing backlash and having reliablility. Bearing adjustment gave me almost 70 oz-in
improvement.

 For any given stepper what you are after is max power which is torque times speed and that occurs in rather narrow range.
Go over that max which is required for some machining task and you have an unreliable situation.

Just some opinions,  
RICH