• One (1) spurious whitespace.
• One (1) occurrence of "random(3) bad, arc4random(3)" good.
• Three (3) writes that will never be seen.

The latter two points are complaints from clang-analyze. Switching to
arc4random(3) is decidedly a good idea because we weren't doing any kind
of PRNG seeding anyways. The discarded assignments are arguably good
for future-proofing, but it's better to improve the S/N ratio from
clang-analyze.