We set out to determine the ratio, q IR , of rest-frame 8-1000-μm flux, S IR , to monochromatic radio flux, S 1.4 GHz , for galaxies selected at far-infrared (IR) and radio wavelengths, to search for signs that the ratio evolves with redshift, luminosity or dust temperature, T d , and to identify any far-IR-bright outliers -useful laboratories for exploring why the far-IR/radio correlation (FIRRC) is generally so tight when the prevailing theory suggests variations are almost inevitable. We use flux-limited 250-μm and 1.4-GHz samples, obtained using Herschel and the Very Large Array (VLA) in GOODS-North (-N). We determine bolometric IR output using ten bands spanning λ obs = 24−1250 μm, exploiting data from PACS and SPIRE (PEP; HerMES), as well as Spitzer, SCUBA, AzTEC and MAMBO. We also explore the properties of an L IR -matched sample, designed to reveal evolution of q IR with redshift, spanning log L IR = 11-12 L and z = 0−2, by stacking into the radio and far-IR images. For 1.4-GHz-selected galaxies in GOODS-N, we see tentative evidence of a break in the flux ratio, q IR , at L 1.4 GHz ∼ 10 22.7 W Hz −1 , where active galactic nuclei (AGN) are starting to dominate the radio power density, and of weaker correlations with redshift and T d . From our 250-μm-selected sample we identify a small number of far-IR-bright outliers, and see trends of q IR with L 1.4 GHz , L IR , T d and redshift, noting that some of these are inter-related. For our L IR -matched sample, there is no evidence that q IR changes significantly as we move back into the epoch of galaxy formation: we find q IR ∝ (1 + z) γ , where γ = −0.04 ± 0.03 at z = 0 − 2; however, discounting the least reliable data at z < 0.5 we find γ = −0.26 ± 0.07, modest evolution which may be related to the radio background seen by ARCADE 2, perhaps driven by <10-μJy radio activity amongst ordinary star-forming galaxies at z > 1.