Can you write down the specific dependency structure between the priors you’re trying to achieve? From the code, it looks like you have something like \beta_{i} \sim \text{LogNormal}(\log(\beta_{i-1}), \sigma^2), is that your intent?
I ask because the code you use to generate data doesn’t appear to have this dependency between tasks.
To debug your specific code, I’d need to see the traceback of the error that you’re getting is. I can say that you need to use the .dist method to create random variables in the scan inner function, so the coef_prior line will raise an error.