In the other kata (http://www.codewars.com/kata/5508249a98b3234f420000fb) it is said:

if possible the message will be evenly split between the five runners; if not possible, parts 1, 2, 3, 4 will be longer
and part 5 shorter. The fifth part can have length equal to the other ones or shorter. If there are many options of how > to split, choose the option where the fifth part has the longest length.

For example, if the coded message has a length of 17 the five parts will have lengths of 4, 4, 4, 4, 1. The parts 1, 2, > 3, 4 are evenly split and the last part of length 1 is shorter. If the length is 16 the parts will be 4, 4, 4, 4, 0.
Parts 1, 2, 3, 4 are evenly split and the fifth runner will stay at home.

I hope this helps you. Tell me if this works!

Unless you would want the distribution of 257 letters between 5 messengers to be 52, 52, 52, 52, 49? I suppose it makes sense in the problem description, but it's definitely not optimized.

Imagine 320 messengers carrying 2568 letters. My distribution would be 9 for the first 8 messengers and 8 for the rest. The distribution I described above won't work at all.

Reload the page, I have spoiler flagged my answer but you must able to see both!

You are welcome!

I found a bug from my code.
Thank you for your reply.

The test cases are right, it seems that you have an error in your encoding, you may have forgotten to encode a terminal ";" :-(
Before thinking that tests have errors, ask yourself if your code is good:-)

in java.
encoded string doesn't have a ';' in "sft b-dspxejoh". but test case expected
"expected:<...he shores a-crowding[;]> but was:<...he shores a-crowding[]>"

Which language? Why do you think there are errors? Can you be more explicit please?

I think test case (3, 6, 7) may contains some error.