Why is the author sure we need microbes to discover exploits in a simulation: the "floating point error" problem would be far more obvious, like being unable t

Why is the author sure we need microbes to discover exploits in a simulation: the "floating point error" problem would be far more obvious, like being unable to measure sub-planck lengths or model singularities?