I have added some boldface
A simulation tells you that a given circuit might work. A prototype tells you that your particular implementation and construction of that circuit appears to work.

any board, however it is made, production board or prototype tells you nothing about any circuit other than the one immediately in front of you.
It tells you that the components might work. It tells you that the configuration might work.

When events involve a few MHz, or even oe or two tens of MHz, a hand-built prototype, fabricated in the kitchen or basement, can behave quite well
here we differ. I doubt that the frequency is the problem, but rather think it is the rise and fall times of the signals. Of course, if you slow down to a crawl, the ringing may die down before it has done any harm.

and now for the old adage:
succesful testing does not prove that there are no bugs, it only ptove that there are no KNOWN bugs

examples of "succesful testing of defrective products":
Hardware: We have some purchased boards where some transistors typically fail after 5-7 years of service. When new they all "test" beautifully and without showing any signs of failure. Investigation has shown that these transistors are "driven to the hilt" (100mA base current to a SOT-23 rated at max 500mA pulsed at 1A) and, thus are stressed enough to make early failure likely; however they "test" beautifully when new.
Software: I had the opportunity to consult on a "perfectly working, fully tested" software that was running on many units. The reason for bringing me in was that "with 1000 boards running one of them (not the same every time) will lock up about once a month". This turned out to be due to "if interrupt b happens during interrupt a which happenrd while the main was executing lines x through y, the board locks.

Now, gentlemen, tell me how "testing" would have revealed either of the above.

Erik