Why does time.sleep in unrelated code location affect time.perf_counter_ns?

I am trying to use time.perf_counter_ns to time how long something takes in my code. It all works as expected if the code looks like this:

import time
total = 0
num_samples = 1000

for i in range(num_samples):
    s = time.perf_counter_ns()
    e = time.perf_counter_ns()
    total += e - s
print(total/num_samples)

$ python3 test.py 
91.507

However, when I add a time.sleep(1) in an unrelated spot in the code, the time I am measuring is higher:

import time
total = 0
num_samples = 1000
time.sleep(1)     # <<< I didnt think this should matter!!
for i in range(num_samples):
    s = time.perf_counter_ns()
    e = time.perf_counter_ns()
    total += e - s
print(total/num_samples)

$ python3 test.py 
417.304

So, why does adding a time.sleep(1) in a seemingly unrelated spot in the code makes a difference?

Using:

Python 3.8.10 (default, Sep 28 2021, 16:10:42) 
[GCC 9.3.0] on linux