If I see a statistic of homicide-related deaths per 100,000 is that per 100,000 deaths therefore giving an expected probability of death when dividing one by the other? or is it deaths by homicide per 100,000 people per period of time (like annually or something) giving a smaller figure because of all the people who didn't die?
It is per 100,000 people in the population per period of time, not per 100,000 deaths.
In the United States in 2016 there were 17,250 murders, 2,744,248 deaths, and ~322,747,241 people, with an officially-reported rate per 100,000 people of 5.3. If it was murders per deaths the rate would be (17,250*100,000/2,744,248) = 629. Calculating it per 100,000 people (17,250*100,000/322,747,241) gives the same number as reported by the FBI, 5.3.