Signal and noise have no relation at all to grain in the context of silver halide films. The only noise is fog: unexposed grains that develop anyway. Everything else is "signal". The size and shape of the resulting grain is dependent on the type of halides, the shape of the grain, and the required speed when formulating the emulsion.
Your developer will not change the actual grain size. That is built into the emulsion when it is produced. No developer, solvent or not, is going to break up the grain into smaller pieces. But size and shape do have to do with the developer and how that either encourages or discourages the growth of filamentary silver. This is quite dependent on the pH at which the developing agent is active, and the type of developing agent. It's not actually clear in later research that there is even a strong relationship between solvency and grain shape or size. There were suggestions for a long time that physical development played a big role and solvency was required to further that. But later research and developer formulations show that even this is not well understood, despite earlier works touting it as fact.
What is clear that p-aminophenol developers have a certain set of characteristics that people either like or don't. Tweaks to concentration, pH, and buffering have an effect on the output, hence the differences between Rodinal and R09. But it's all within a narrow range of results. p-aminophenol is only usefully active in a certain pH range (like all agents, though these ranges differ widely) and only stores well in a certain concentration and thus at a certain pH. Those two facts limit the potential results to a narrow spectrum. Hence the similarity of all of these developers.