Context. The properties of solar flare plasma can be determined from the observation of optically thin lines. The emitting ion distribution determines the shape of the spectral line profile, with an isothermal Maxwellian ion distribution producing a Gaussian profile. Non-Gaussian line profiles may indicate more complex ion distributions. Aims. We investigate the possibility of determining flare-accelerated non-thermal ion and/or plasma velocity distributions. Methods. We study EUV spectral lines produced during a flare SOL2013-05-15T01:45 using the Hinode EUV Imaging Spectrometer (EIS). The flare is located close to the eastern solar limb with an extended loop structure, allowing the different flare features: ribbons, hard X-ray (HXR) footpoints and the loop-top source to be clearly observed in UV, EUV and X-rays. EUV line spectroscopy is performed in seven different regions covering the flare. We study the line profiles of the isolated and unblended Fe XVI lines (λ262.9760 Å) mainly formed at temperatures of ~2 to 4 MK. Suitable Fe XVI line profiles at one time close to the peak soft X-ray emission and free of directed mass motions are examined using: 1. a higher moments analysis, 2. Gaussian fitting, and 3. by fitting a kappa distribution line profile convolved with a Gaussian to account for the EIS instrumental profile. Results. Fe XVI line profiles in the flaring loop-top, HXR footpoint and ribbon regions can be confidently fitted with a kappa line profile with an extra variable κ, giving low, non-thermal κ values between 2 and 3.3. An independent higher moments analysis also finds that many of the spectral line kurtosis values are higher than the Gaussian value of 3, even with the presence of a broad Gaussian instrumental profile. Conclusions. A flare-accelerated non-thermal ion population could account for both the observed non-Gaussian line profiles, and for the Fe XVI "excess" broadening found from Gaussian fitting, if the emitting ions are interacting with a thermalised ∼ 4 MK electron population, and the instrumental profile is well-approximated by a Gaussian profile.