Every time something isn't behaving as it should, somewhere in the bytes of RAM will be the reason it isn't working properly. You could dump that ram and single step the code to find the fault. And then fix it for the millions of other people out there.
Or you could just reboot, and the problem will be erased, and probably never happen again to you for many years, by which time you'll have moved onto new hardware anyway.
1) a lot of cancer research occurred between the random bacteria observation and viable chemotherapy.
2) ignoring the above, you still have too compare the success rate of funding cancer research vs random experiments.
If I'm remembering correctly, the scientists were trying to find a way to see if alcohol impacted some drug, but wanted the measurement to be blind. So don't participants would need to drink alcohol without being sure if they had or not. The author and his wife spent a lively evening trying to find some kind of mix that hid the alcohol and found grapefruit to be very good at it.
Then when the actual study ran, they had participants take the drug or a control as well as vodka with grapefruit or just grapefruit. And to their surprise, the grapefruit had a massive effect regardless of the vodka!
"But vindication implies that we are out to get somebody. And we weren't. We had no idea that this was out there."
https://www.npr.org/2015/09/24/443053672/how-a-little-lab-in...
Kudos to the NC state team for setting up this control, and digging into the results, especially when the purpose of their experiments was different (their intent was to find a better way to deliver pesticides).
It is amazing how many times this happens - unintended or unanticipated results from an experiment, often peripheral results, which turn out to be extraordinarily interesting (usually for a different purpose as the original experiment), if observed and followed-up on with rigor and innovation.