>You present strawman (seeing it makes it credible)
Seeing something happen does make an allegation more credible; this is the whole point of, say, witness testimony. Why we are more likely to believe things which are attested to by others. Credibility is also an element of character and relation to the victim. If I told my parents I was mugged in an underpass on my way home, which explains why I don't have my cell phone, I'd be a little weirded out if they simply replied with "your allegations are not credible, that's a strawman."
>we can and should considered guilty until proven innocent.
Guilt and innocence in which epistemic context? Consider that there are both very well known cases of public opinion being wrong (and overturned by a court or other method), or the findings of courts being wrong (and overturned by public investigation and revelations). Guilt in one arena does not necessarily equal guilt in another[0], and I think that's part of the point GP was trying to make.
>a couple fights, she takes the kids, he shouts, she calls the police and claims abuse, he is beaten and arrested; during the divorce she uses the arrest as evidence of the abuse, and frames the beating as evidence of her husbands violent tendencies
You can't seriously accuse GP of setting up a straw-man, then relay an idealized (for your point) situation which is, to my knowledge, not backed up by any statistical power.
>Because this is what I'm arguing against, and I can't understand how or why anyone would want this to be how things work.
You're arguing from a view-from-nowhere, that is, you have already relayed the facts of an unfair situation and the legal consequences of that situation, without taking into account that the situation you relayed is a failure of the justice system, not only of the allegation. Your depiction of how a single allegation snowballs like that isn't impossible - but you haven't done anything to show that this is the result of most (or even a significant number of) allegations of domestic abuse.
[0] "In at least a wide variety of cases, our orientation toward testimony is a non-skeptical one: when someone tells us that P, we end up justified in believing
that P. And ordinarily, if the speaker knows that P and tells us that P, we end up ourselves knowing that P. That testimonial knowledge is common is both commonsense and philosophical orthodoxy.
But we do not always take this kind of stance. Sometimes we interrogate testimony; sometimes we explain it away or demand proof. As much recent literature in both social psychology and epistemic injustice has shown, social and cultural factors influence the degree to which testimony is considered credible."