Not true. First off, we DO NOT KNOW if these treatments will work.
We have reason to believe the treatment will work. We also have reason to believe they're safer than death (we have reason to believe they're reasonably harmless in their own right, but that's extreme compared to the benchmark of preventing death). Thus your statement is both misleading (not entirely inaccurate) and irrelevant.
Not true. That's the point -- more statistical data is useless if you're not collecting it in a way that will give you an answer. The only way to get an answer is with a randomized, controlled trial. Animals aren't humans.
I see you've done no real statistics or explored any real science.
Many mechanisms of actions for many modern drugs are explained by animal models. For example: anything that affects brain chemistry is explained by experimentation on rats. We know exercise and noopept (nootropic drug) both improve learning dramatically by increasing BNF and BDNF levels in the brain--because we tried this with rats (run on a wheel, then run a maze; compare to a lazy rat, and the exercised ones learn twice as fast, consistently) and then cut their heads open to see what was going on in their brains (this is fatal). Anti-depressants, SSRIs, NDRIs, and other drugs that modify your brain chemistry directly are known to work in those specific ways by animal model--nobody actually checked an actual human, ever.
And just in case you say, "Well, it works half the time" -- the database studies they do are much more thorough than anything they could do in west Africa, where they don't even have medical records or death certificates. They do these studies in places like Sweden, where they have detailed medical databases of every citizen from birth to death. We don't have data like this for west Africa.
And this is where you show you don't understand what you're talking about.
"It works half the time" is roughly chance. "It's wrong 99% of the time" is procedure and interpretation: you're reading the data wrong, getting consistently inverted results, and resolving this will give you 99% accuracy. Statistics isn't a right-to-wrong scale; it's a consistent-versus-inconsistent scale, and so both always-right and always-wrong are valuable and meaningful, and less-consistent is less-meaningful.
It turned out that the nurses who took hormone replacement therapy were different from the nurses who didn't -- they were more concerned about their health, so they exercised more, controlled their weight more -- and took hormone replacement therapy, because they thought it might be healthy.
Yes, and we also have some facilities with poor handling procedure and good handling procedure. In our case, however, there's ebola turn-over in these facilities. To be blunt: facilities with poor handling procedure will show lots of cases regularly of ebola contraction amongst healthcare workers; if we vaccinate their workers and the rate of infection drops off, we have a self-controlling experiment.
Such a self-controlling experiment suggests, with confidence, that the vaccine works. We'd need more long-term data to discover the vaccine's side-effects (if any), as well as long-term human blood work study and more invasive (dangerous, harmful) animal model study and in-vitro study to further understand and improve the mechanism; but that's true in any case. Vaccinating half the population and giving the other half a saline placebo won't get you any more insight in those areas.