Summary

As discussing in the simulation result, we've update the specification to have two tolerance parameters $\lfloor\delta_{EEF}\rfloor$ and $\lceil \delta_{EEF} \rceil$ instead of just $\delta_{EEF}$. It's for applying different tolerances to malicious peer detection and unhealthy peer detection. The same applies to $\delta_{DEF}$ as well.

Reason

We've found that we may need to set $\delta_{EEF}=1.00$ to avoid marking healthy connections as malicious, as investigating the simulation result. But then, no peer would be detected as unhealthy in the end, if $EEF=1$, because $EEF - (EEF \times \delta_{EEF})=0$. Of course, if a peer emits no message at all, it will be detected as unhealthy. However, we want to make the unhealthy tolerance more useful. The conclusion is separating the unhealthy tolerance $\lfloor\delta_{EEF}\rfloor$ from the malicious tolerance $\lceil \delta_{EEF} \rceil$.