Is it worth it to hand-test code when doing an online assessment without an interviewer watching?

Sometimes it takes a while to write out the full trace of a test case to check the algorithm works as expected.

Do most platforms have a difference between "run code" and "submit" like leetcode does? Do you get penalized for using "run code" loads of times on your own testcases if they fail a lot?