Strange, that article in National Review.  It conflates unrelated
factoids to serve a predetermined thesis.

  https://www.nationalreview.com/2018/05/maurizio-pollini-pianist-beautiful-imperfection-carnegie-hall/ 

The roll industry arranged rolls from sheet music since the very start
of the trade in the 1870s.  And most everything was done that way right
through.

Aeolian (and many others) introduced hand-played rolls in 1910, and at
the start they just issued the results, remarkably rough at times.  The
craze for dance music exploded about 1916/7 and required greater timing
precision -- so dance rolls were made progressively tidier up to 1919.
(You can follow this evolution looking at original Gershwin rolls.)
Dance rolls evolved into a subtle form of arranged product.

But, classical music was never done that way!  Perforating uses
fixed-length steps between perforations: smaller steps allow more
precise timing (like pixels in a digital image). Unfortunately, earlier
hand-played Aeolian and Ampico rolls used large steps -- around 20 rows
per inch along the roll, which was rather coarse.  (Music critics
complained at the time, and do so now.)

Some classical rolls were somewhat tidied to compensate, but more
usefully production standards were improved by both companies to 30
rows/inch for some rolls.  By comparison, Hupfeld used 48 rows/inch,
and Welte 63 rows/inch (in its T98 'Green' format but identical
precision on T100 'Red' rolls) -- the resulting timing smoothness
has always greatly helped Welte's musical reputation.

So, this was an interesting conundrum: coarse production standards
forced dance rolls to be quantised.  If rolls had been manufactured to
higher timing precision to start, quantisation may never have evolved.
And we can only speculate how reputations might differ if Aeolian and
Ampico had matched Welte's timing accuracy.

But the argument that inaccuracy leads to an improvement is surely
false, especially when it conflates differing types of inaccuracy --
a great artist intelligently adapting to limitations is simply not
a machine dumbly making random errors.

And arguments that artists are "too precise" is just a way of saying
you don't like their artistic decisions.  Nobody would ever advance
similar arguments about, say, surgeons or scientists.

Julian Dyer