Context. Ultracompact X-ray binaries (UCXBs) typically consist of a white dwarf donor and a neutron star or black hole accretor. The evolution of UCXBs and very low mass ratio binaries in general is poorly understood. Aims. We investigate the evolution of UCXBs in order to learn for which mass ratios and accretor types these systems can exist, and if they do, what are their orbital and neutron star spin periods, mass transfer rates and evolutionary timescales. Methods. For different assumptions concerning accretion disk behavior we calculate for which system parameters dynamical instability, thermal-viscous disk instability or the propeller effect emerge. Results. At the onset of mass transfer, the survival of the UCXB is determined by how efficiently the accretor can eject matter in the case of a super-Eddington mass transfer rate. At later times, the evolution of systems strongly depends on the binary's capacity to return angular momentum from the disk to the orbit. We find that this feedback mechanism most likely remains effective. In the case of steady mass transfer, the propeller effect can stop accretion onto recycled neutron stars completely at a sufficiently low mass transfer rate, based on energy considerations. However, mass transfer will likely be non-steady because disk instability allows for accretion of some of the transferred matter. Together, the propeller effect and disk instability cause the low mass ratio UCXBs to be visible a small fraction of the time at most, thereby explaining the lack of observations of such systems. Conclusions. Most likely UCXBs avoid late-time dynamically unstable mass transfer and continue to evolve as the age of the Universe allows. This implies the existence of a large population of low mass ratio binaries with orbital periods ~ 70 - 80 min, unless some other mechanism has destroyed these binaries.