nithp | - View NITH result

You have discovered one of the big issues involved with using any sort of sleep function to manage future events. The way asyncio.sleep works is that it registers a future. A future is like a placeholder for a function that hasn't returned yet. You don't have to wait for the function until you actually want to read the result but the function is still processing in the background. Futures are one of the foundational elements of asynchronous processing.

Technically asyncio.sleep should work in this context. I'm not sure entirely what is happening but most likely the discord.py library has some sort of manager that drops these futures if they're suspended for more than a certain amount of time. If I find out more about this then I'll edit this answer.

The long and the short of it is that you should not use sleep calls for long events like this one. Instead, you should create a global (or guild-level) database and have a looping task that reads from that database and calls a function that runs all the code below the asyncio.sleep. You could do this by creating a dictionary that stores when a users nitro will finish along with some other information like the author and channel.

You can find out more about tasks in the discord.py library here