Is then the trial and error method, the only possible way to find out the optimum grid solution!! - so that a balance between accuracy and performance can be achieved for the modeling process.
Grid size for the watershed should be based on the size of your study area and purpose of the study. Nowadays DEMs are available at high resolution. If your watershed grid size is closer to DEM resolution, there will be too many computational points. This will create a problem for large watersheds. In addition, the purpose of the study is important in deciding your grid size. In my experience, if you are analyzing water quantity or movement smaller grids will bring only marginal improvement. However, if your analysis is on water quality it might bring some noticeable improvement for better resolution grids. In summary, the size of the study area and purpose should decide your watershed grid size. Note that you can strike a balance between quality of results and grid size within a few trials. All the best
In my opinion the grid generated for the water shed should be at least equal to the grid size of the DEM resolution i.e., if the DEM is 2mx2m, then the grid size for water shed should be same size or more refined if possible for better outputs.