We study the dependence of fragmentation in massive gas-rich galaxy disks at $z > 1$ on feedback model and hydrodynamical method, employing the GASOLINE2 SPH code and the lagrangian mesh-less code GIZMO in finite mass mode. We compare non-cosmological galaxy disk runs with standard blastwave supernovae (SN)feedback, which introduces delayed cooling in order to drive winds, and runs with the new superbubble SN feedback, which produces winds naturally by modelling the detailed physics of SN-driven bubbles and leads to efficient self-regulation of star formation. We find that, with blastwave feedback, massive star forming clumps form in comparable number and with very similar masses in GASOLINE2 and GIZMO. The typical masses are in the range $10^7-10^8 M_{\odot}$, lower than in most previous works, while giant clumps with masses above $10^9 M_{\odot}$ are exceedingly rare. With superbubble feedback, instead, massive bound star forming clumps do not form because galaxies never undergo a phase of violent disk instability. Only sporadic, unbound star forming overdensities lasting only a few tens of Myr can arise that are triggered by perturbations of massive satellite companions. We conclude that there is a severe tension between explaining massive star forming clumps observed at $z > 1$ primarily as the result of disk fragmentation driven by gravitational instability and the prevailing view of feedback-regulated galaxy formation. The link between disk stability and star formation efficiency should thus be regarded as a key testing ground for galaxy formation theory.

Using an isolated Milky Way-mass galaxy simulation, we compare results from nine state-of-the-art gravito-hydrodynamics codes widely used in the numerical community. We utilize the infrastructure we have built for the AGORA High-resolution Galaxy Simulations Comparison Project. This includes the common disk initial conditions, common physics models (e.g., radiative cooling and UV background by the standardized package Grackle) and common analysis toolkit yt, all of which are publicly available. Subgrid physics models such as Jeans pressure floor, star formation, supernova feedback energy, and metal production are carefully constrained across code platforms. With numerical accuracy that resolves the disk scale height, we find that the codes overall agree well with one another in many dimensions including: gas and stellar surface densities, rotation curves, velocity dispersions, density and temperature distribution functions, disk vertical heights, stellar clumps, star formation rates, and Kennicutt-Schmidt relations. Quantities such as velocity dispersions are very robust (agreement within a few tens of percent at all radii) while measures like newly formed stellar clump mass functions show more significant variation (difference by up to a factor of ∼3). Systematic differences exist, for example, between mesh-based and particle-based codes in the low-density region, and between more diffusive and less diffusive schemes in the high-density tail of the density distribution. Yet intrinsic code differences are generally small compared to the variations in numerical implementations of the common subgrid physics such as supernova feedback. Our experiment reassures that, if adequately designed in accordance with our proposed common parameters, results of a modern high-resolution galaxy formation simulation are more sensitive to input physics than to intrinsic differences in numerical schemes.