I see in some answers from 2014, some talk about independent sources of entropy mixed canonically with XOR with say one source which is known in advance.
Assuming we are operating on bit string lengths of length $m$: an entropy generator $G$ would yield $s1 ^ s2 ^...S_n-1 ^ 1 = Z[output]$. Here we have set our known bit-string to true, i.e. $S_n = 1111..1(m times)$
It is accepted that, being independent sources, the one 'bad' source doesn't make the seed any worse. However, by XORing $Z$ with $S_n$ gives us the $S_1..n-1$, so we know the entire state yielding $Z$.
Why is this not bad? I might be missing something very fundamental here but can this knowledge not be exploited knowing the output?