A drop of water falling on a hill does not wash it away.
So, if we start with a hill, then after a drop of water we still have a hill.
After another drop, we still have a hill; and many repeated applications
of the first, italicised line means that after lots of drops the hill remains.
But, after enough drops the hill will, of course, have been eroded away.
That is basically a Sorites paradox. Similarly, all real-world calculations will, if long enough, become swamped by error bounds. All measurements should come with error bounds, and while a short calculation will result in only slightly larger error bounds on the result, a very long calculation will be useless. Now, logic is supposed to be different, more like Geometry, where given certain lengths, geometrical manipulations can be arbitrarily long. But that will only be the case if the terms that the logic is applying to are definite. In the real world, there is a ubiquitous, if usually very slight, vagueness (it is there because it is so slight: nothing has acted to remove it). Consequently logical arguments that are about real things should not be too long. It is an interesting question, how long they can be; but certainly, those of the Sorites paradoxes are too long.