Whenever a scenario is run through the model Path creates what is called a "results scenario" at the time of the run. By default this new scenario contains copy of all of the original model inputs – Path merges all of the input dependent scenarios into this results scenario, makes the results scenario read-only, and places this "flattened" copy of the model inputs in the database along with the output. This way the results scenario always contains an accurate record of exactly which inputs were used at the time of the model run, even if the user were to subsequently modify one of the original input scenarios after the run.
Of course the downside of this option is that you create copies of every model input each time you do a run, which can take up more processing time (and presumably more time for writing to the database). You can choose to uncheck this option for any of the input scenarios – either the parent scenario your are running, or any of its dependent scenarios. Doing so tells Path not to make a copy of those inputs into the results scenario at runtime, but rather to continue representing this scenario as a dependency at runtime. This can save database space, particularly for scenarios that contain lots of inputs. The downside of this approach is that if you were ever to edit one of your original input scenarios, you would lose your record of exactly which inputs were used to generatae any previous run results.
I imagine you would get better performance if you unchecked this box for your scenarios – which would be acceptable if you don’t ever plan to subsquently edit the original input scenarios after the runs are complete. I’d be interested to hear how much improvement in performance you see over the default option if you’re able to run a short test.