In embedded 3D printing (EMB3D), a nozzle extrudes continuous filaments inside of a viscoelastic support bath. Compared to other extrusion processes, EMB3D enables softer structures and print paths that conform better to the shape of the part, allowing for complex structures such as tissues and organs. However, strategies for high-quality dimensional accuracy and mechanical properties remain undocumented in EMB3D. This work uses computational fluid dynamics simulations in OpenFOAM to probe the underlying physics behind two processes: deformation of the printed part due to nearby nozzle motion and fusion between neighboring filaments during printing. Through simulations, we disentangle yielding from viscous dissipation, and we isolate interfacial tension effects from rheology effects, which are difficult to separate in experiments. Critically, these simulations find that disturbance and fusion are controlled by the flow of support fluid around the nozzle. To avoid part deformation, the nozzle must remain far from existing parts during non-printing moves, moreso when traveling next to the part than above the part and especially when the interfacial tension between the ink and support is non-zero. Additionally, because support can become trapped between filaments at zero interfacial tension, the spacing between filaments must be tight enough to produce over-printing, or printing too much material for the designed space. In non-Newtonian fluids, spacings for vertical walls must be even tighter than spacings for horizontal planes. At these spacings, printing a new filament sometimes creates and sometimes mitigates shape defects in the old filament. While non-zero ink-support interfacial tensions produce better inter-filament fusion than zero interfacial tension, interfacial tension also produces shape defects. Slicing algorithms that consider these unique EMB3D defects are needed to improve mechanical properties and dimensional accuracy of bioprinted constructs.