The memory request is received by another AI, which can decide to reject something that doesn’t follow a typical pattern of “things about the user”. Then it has to decide to store a new memory, or has the ability to replace or delete them by the message that ChatGPT sends to it.
That means your multi-paragraph might have just gotten a judgemental “nope” you never hear about.
You also may need to refresh the browser to actually see the current state of memories or custom instructions. The settings often don’t seem to populate the user dialogs when you expect them. Turn the memory feature on and do a refresh just for good measure that the setting is “taken”.
Well, it seems that by trying to demonstrate this, with a bit of overly-direct instruction, I just turned down the IQ of ChatGPT…
No, I didn’t purge all my existing fun facts stored about me just for a demonstration. I leave it disabled - it is that uninteresting.