I did pumps and motors for about 5 years at my first real machinist job (have had a home shop and worked in an aviation museum as a restoration machinist for almost 15years before that), left for four years and worked in a chrome plating/hydraulic shop and went back to the pump shop last October.
Modern digital computer balance systems like the one I use have virtual key compensation. You just input the size of the keyway, diameter of shaft, distance from the machine roller to center of key, round ends or not (open ended impeller keyway), and it calculates out the error. Just started with the balance machine first of the year, but it's not too bad so far. Company that sold us the machine calibrates it every year. Have not had it a year, so we will see how far out it is first of next year.
Not sure what kind of metal spray rig you are running. I'm running a Metco wire gun. Doesn't get the work nearly as hot as a powder spray rig. If you have a lot of repetitive work, you can save a LOT of time spraying by spraying one and turning it to exact size, polish to fit, etc... while hot. Let it cool and you may have to do it again, as it will shrink, of course. Mic it cold and see how much it shrank. Add the amount undersize it came out to the original measurement when hot and now you have a constant to take that size bearing to immediately after spraying. Leave them that amount over and they will cool to exactly on size. When I started spraying, they were getting two shafts a day done, because they had to let them cool before turning. I can do eight shafts a day now, if they are common sizes I have compensated.
Agree with all the above. On sleeving wallowed bearings, I leave the sleeve only about .050 undersize, at most and then finish bore without disturbing the setup, if at all possible. That gives you one .010-.020 pass to figure out where you are, another .010 to stabilize, a .010 to give an honest cut and the third can be adjusted a few thousandths either way to land the bearing fit. I have five machines I use to bore with, depending on how the work is made. I have a 12" squarehead LeBlond for small lathe work, a 20" American Pacemaker for large lathe work, a Cincy #3 vertical for stuff too big or awkward to get in the lathes, a Cincy #2 horizontal for stuff that works better that way than vertical, and a 42" Webster and Bennett vertical turret lathe for stuff too big for anything else.
DROs are nice, but unless you have a tenth reading scale and display, they are not very useful for finishing a bearing. The VTL has a DRO and it is very useful for getting in close, but it only reads in .001 on the radius (can't get closer that .002 on diameter). Last few passes have to be made with a DI on the rail or using the handwheel dials, which can be visibly split into .0025 increments. Most anything big enough to go in that machine will have a bearing housing tolerance of .0007 or so... you can work with that.
The lathes both have crossfeed and compound dials graduated in diameter, put .010 on the dial, it takes .010 off the diameter. Swing the compound to 60 degrees and on a radius reading dial, it will now take half as much, so .010 on the compound means it takes .010 off the diameter. With the diameter reading dials, it goes to half that, so putting .010 in on the dial means it takes .005 off. Each .001 graduation takes .0005 off. Again, eyeballing it to quarters between the division will land you at .00012. You can't skim a .00012 cut, but you can leave the dial under or over by a quarter of the division of the dial and land fits within .0002 pretty danged easily as long as your cuts are the same depth, same speed, same feed, same tool, etc... in other words, DON'T CHANGE ANYTHING on those last three .010 passes to maintain stability.
For boring on the Cincy mills, I have a big Chandler Duplex boring/facing head. The dial on it is graduated in .0005 increments on diameter, so again, you can hit .0003-.0005 pretty easily on those final passes.