Volonteri et al. (2011) found that the number of radio-loud quasars above redshift 4 calculated from the luminosity function (based upon Swift/BAT observations) is much smaller than the number estimated from the known high-redshift beamed sources, blazars, assuming that for every beamed source with a Lorentz factor of $\Gamma$, statistically $2 \Gamma^2$ non-beamed sources should exist. To explain the missing misaligned (non-beamed) population of high-redshift sources, they proposed various explanations, involving heavy optical obscuration and significantly different Lorentz factors at early cosmological epochs. Our EVN observations targeting high-redshift ($z>4$) blazar candidates revealed 3 sources not showing relativistic beaming, but rather kpc-scale double structures. These three sources have significant radio emission resolved out with the EVN, while they are compact on $\sim 5-10$ arcsec scale. Our dual-frequency ($1.5$ and $5$ GHz) e-MERLIN observations of these three sources revealed a rich morphology, bending jets, and hot spots with possible sites of interaction between the jets and the surrounding medium at intermediate scales.