Very large participant such as MrKermit who moved from project to project have a considerable temporary impact on what science gets the momentary bigger chunk, but WCG indeed balances through the pool of multi-project participants how fast each project goes. Eventually though if for example members representing 20 percent of the grid power tick only HFCC and disable the 'if there is no work', I think it would be counter for WCG to insist on the even share principle if the work units are available of course... you thus see that HCMD, HFCC, FAAH and RICE can and are moving faster than others.

Below a chart on the balance of power over the last ~100 days. It shows that recently RICE was taken out of forced slow mode into normal share, shifting the powers. Uplinger explained this tuning happens continuously also because the work units are not always equally hard to compute or not unlimited in daily supply. Given the grid is in perpetual growth some sciences pick up without much impacting the others.

Sort of how I understand it works.