Old Robot for milling

The issue with buying any old robot is that the older a robot is, the more likely it is to have hardware issues, and spare parts and service are harder and more expensive to obtain.

Also, a used robot may be in poor condition -- even if 100% functional, it may suffer from poor accuracy, increased backlash, etc. This can be overcome for basic operations like spot welding, but for high-accuracy applications like milling (which are already difficult for brand-new robots) it becomes an even greater problem.

Using any robot for milling operations will require a special CAD/CAM tool chain. Either a special robot post-processor to convert the milling program into robot program, or a great deal of manual manipulation. The software is not cheap, so you will probably need to become an expert in both G-Code and the robot program, and create a means to convert from one to the other.

"Streaming" program locations to a robot is... possible... but not simple. The best approach is probably to "segment" the program, and create a structure where the robot would complete one segment, move to a clear position, then download all the points for the next segment, before beginning the segment. This is slower, but less risky than attempting to stream "real time" data to the robot.

I do not think any particular brand of robot is particularly best for milling. Personally, I prefer KUKAs -- I think they have the best path planning, physical robustness, and are the most "hackable." One should avoid buying a used Fanuc at all costs -- they're not bad robots, but the company will refuse to sell you service or parts unless you pay them $10000.00 to "re-license" the operating system first.

What do you want to mill? even if your spindle is light you must not forget that the milling itself will put force on the robot, but this is very depended on what you want to mill. Also in my experience a heavier robot helps with rigidity and thus in your milling quality.