This avoids the stat(2) of /etc/localtime per message scan if TZ isn't
set in the environment; that's the default case on this Linux system.
I can't see a way the TZ environment variable is likely to change
between messages, or that it's intended an invocation of scan(1), say,
copes with the timezone shifting underneath it by /etc/localtime being
changed.