Concern on the size of normalized code, compiling complexity.

I'm not quite familiar with compiling theory, category theory and many other related topics. But I'm very interested in all of them.
Within my limited knowledge, I'm concerned about two things.
- I think the normalized code's would be unnecessarily big when, for example, it de-sugaring a program which has a constant number as a recursion time.
- When the source code contains a constant which needs O(2^n) time to compute, the compiling time will be at least O(2^n).

How can you handle these two problems, if I'm not asking stupid questions? Thank you for your attention.

No files, tests, or entry points are named. First read the normalization and compilation implementation and existing tests, then determine whether the two complexity concerns are reproducible and establish a settled handling strategy before proposing any change.

Written by the indexing model from the issue text.