It's not a real bug but very inconvenient.
When I run a REMD simulation, I could use
thermo_modify flush yes
to force the output of thermo data to log files.
However, this seems only to work with the slave log files
not the master log. The master log is always cached on
the server I'm using until the job has finished. But
if the job is killed due to any reason (e.g. overtime), I cannot
get the master log. This means the data obtained are
wasted. Could you please add fflush call for the master log?