I'm not sure what that has to do with baby-stepping.
Here is my routine before every print job unless I'm going directly from one print to the next.
1) Get it all heated up.
2) Perform a set of G30s for the autoleveling. (at some point in the past i manually determined the tilt errrors to get the H values).
3) Go to Z1 then drop manually until just grazing a piece of paper. If I'm printing below .2 layers I then lower the head by another few moves of .01 to compensate the thickness of the paper (somewhat). Then I send G92 Z0. But when my print is done or any other time I am forced to re-home, I lose that correction. If I have to add a piece of blue-tape for example, I have to do the same because I have to run the autolevel WITHOUT the tape, then rezero to correct for the change in height (the tape is usually a smaller patch in the middle of the bed that would not cover all of the autolevel probe points).
I've tried measuring the probe/nozzle offset and using that as my auto-level correction but when I change nozzles (for example), that offset changes. I found it easier (and ultimately safer) to just manually do the fine correction to make sure the nozzle is perfect with the bed.
I thought after reading that I might use the baby-stepping to do the exact same thing but without losing that final correction if rehomed (as a post-print command, for example).
As a last thought, does the G30 S1 result survive a re-home?
thanks