Ultracold potassium is an interesting candidate for quantum technology applications and fundamental research as it allows controlling intra-atomic interactions via low-field magnetic Feshbach resonances. However, the realization of high-flux sources of Bose-Einstein condensates remains challenging due to the necessity of optical trapping to use magnetic fields as free parameters. We investigate the production of all-optical K39 Bose-Einstein condensates with different scattering lengths using a Feshbach resonance near 33 G. By tuning the scattering length in a range between 75a0 and 300a0 we demonstrate a tradeoff between evaporation speed and final atom number and decrease our evaporation time by a factor of 5 while approximately doubling the evaporation flux. To this end, we are able to produce fully condensed ensembles with 5.8×104 atoms within 850-ms evaporation time at a scattering length of 232a0 and 1.6×105 atoms within 3.9s at 158a0, respectively. We deploy a numerical model to analyze the flux and atom number scaling with respect to scattering length, identify current limitations, and simulate the optimal performance of our setup. Based on our findings we describe routes towards high-flux sources of ultracold potassium for inertial sensing.