Context. An open question in the study of comets is the so-called cohesion bottleneck, that is, how dust particles detach from the nucleus. Aims. We test whether the CO pressure buildup inside the pebbles of which cometary nuclei consist can overcome this cohesion bottleneck. Methods. A recently developed pebble-diffusion model was applied here to comet C/2017K2 PANSTARRS, assuming a CO-driven activity. Results. (i) The CO-gas pressure inside the pebbles erodes the nucleus into the observed dust, which is composed of refractories, H₂O ice and CO₂ ice. (ii) The CO-driven activity onset occurs up to heliocentric distances of 85 au, depending on the spin orientation of the comet nucleus. (iii) The activity onset observed at ≈26 au suggests a low obliquity of the nucleus spin axis with activity in a polar summer. (iv) At 14 au, the smallest size of the ejected dust is ≈0.1 mm, consistent with observations. (v) The observed dust-loss rate of ≈200 kg s⁻¹ implies a fallout ≥30%, a nucleus surface active area ≥10 km², a CO-gas loss rate ≥10 kg s⁻¹, and a dust-to-gas ratio ≤20. (vi) The CO-driven activity never stops if the average refractory-to-all-ices mass ratio in the nucleus is ≤4.5 for a nucleus all-ices-to-CO mass ratio ≈4, as observed in comets Hale–Bopp and Hyakutake. These results make comet C/2017K2 similar to the Rosetta target comet 67P/Churyumov–Gerasimenko. (vii) The erosion lifetime of cometary planetesimals is a factor 10³ shorter than the timescale of catastrophic collisions. This means that the comets we observe today cannot be products of catastrophic collisions.