We present images of the z = 3.22 radio galaxy 6C 1232 + 39 taken through the B, R, I, J, H, and K filters, an image taken through a narrow-band filter centered on the Lyalpha line, and an image made with the VLA at 5 GHz. Unlike the two other z > 3 radio galaxies that have been extensively studied (B2 0902 + 34 and 4C 41.17), this radio galaxy has a ''classical double '' radio structure. The structure of the radio galaxy is a strong function of wavelength, changing from a long linear structure at the shortest wavelength (B) to one that is dominated by a single compact component at the longest wavelength (K). The galaxy's optical, emission-line, and radio structures are closely aligned; apart from the overall alignment, there is, however, not the clear correspondence between the continuum and Lyalpha structures that would be expected if both trace the same young starburst regions. We show that, like the two other well-studied z > 3 radio galaxies, 6C 1232 + 39 has a red bump in its rest-frame optical spectrum. Low scatter in the K-z diagram is often cited as evidence that radio galaxies formed early in the history of the universe. We show that there is now some evidence that the scatter in the K-z diagram at z > 2 is greater than the scatter at z < 2. We discuss briefly the possible causes of this increased scatter, noting that near-infrared spectroscopy will be crucial in determining whether the increased scatter gives any useful information about the epoch at which radio galaxies formed. Finally we argue that the dispersion in the K-band magnitudes of the known z > 2 radio galaxies, and in particular the existence of luminous objects like 6C 1232 + 39, calls into question the use of the K-z Hubble relation to estimate the redshifts of distant radio galaxies from K magnitudes alone, a method that underpins the only published evidence for a redshift cutoff for radio galaxies.