I'm not sure what method the PPH used to work up their methods, but in general medical and chemical knowledge can be used to build a pretty accurate idea of how a drug will work. By the time a drug hits the market after undergoing the respective clinical trialing, such as through the FDA in the US, there is going to be knowledge of what the drug works on in the body and the mechanism by which is does so. SSRIs, as a random example, work by preventing reuptake of the neurotransmitter serotonin at the synaptic junction between two neurons, allowing the serotonin to have extra time to take effect. Ideally this should increase the benefits of serotonin, but as with any medication can cause side effects as well. Every drug is studied like this.
So to know what drugs are likely to cause death, first you start with knowing how the drug works in the body and use physiology and anatomy to piece together how that could cause problems. Overdose on a beta blocker, for instance, can tank the blood pressure and drop the pulse to the point that the heart is no longer beating effectively enough to live. From there, as you mentioned, you cannot just have people take medication overdoses and see if they die. This is where case studies and statistics come in. Medical professionals love to investigate unique cases. If someone overdoses on a medication that hasn't been studied before due to its lack of commonality, there are good odds the doctor on the case will get permission to make it into a research paper. If it's a common drug to be overdosed on, then you'll find papers on statistical analysis as well as individual case studies.
Take all of this into account and you can put together a pretty good guess of what will kill someone. That is likely how the SN protocol was concocted, if I had to guess. Someone took the physiological process of SN on the human body, realized it would greatly impair oxygen transport, and no oxygen transport means tissue death. It also took into account that absorption occurs most quickly on an empty stomach. They likely then researched cases of accidental SN ingestion and studied the effects. After that they took into account the common side effects of this (nausea, anxiety, tachycardia, etc) and created a plan for how to counteract those symptoms for the least amount of discomfort. Once this plan rolled out and SN became popular, medical journals started pumping out articles analyzing the method from a hospital perspective, because as I said, medical professionals love digging into new cases.
Essentially all of this to say, this is just done through the scientific process. Hypothesize an effect, research it, look at case studies, and ultimately wait for someone to try it and study it after, since research cannot be done on humans. It's possible someone tried these on animal subjects, however I don't know if anyone was willing to go to those lengths or not.
