Error correction is possible in the DNA polymerase, allowing incorrect bases to be removed, and the relevant section redone. When the polymerase recognises an error, it melts the last few bases it attached to the primer strand, allowing it to be translocated to the exo domain.
At the exo domain, the erroneous nucleotide(s) is/are removed, before transferring the strand back to the catalytic fingers, thumb, and palm (synthetic) domains to try again.
Incorrect nucleotides can be recognised by a disruption to the DNA strand backbone (as would occur if two purines or two pyrimidines were bound together), or the phosphodiester bonds form incorrectly (as would happen if uracil is incorporated into a DNA strand). Due to this physical abnormality in the DNA strand, the polymerase is able to detect it as a conformational change, because of the highly specific active site in the enzyme.
The distance between the synthetic and catalytic domains of the DNA polymerase is long. It is not fully understood how the relevant section of DNA is translocated between the two sites. Maybe cryogenic electron microscopy study will improve understanding of this.