The simulations are run with "leap" calendar.
Only 2 "limit.nc" files are used: one for normal (365-day) years, another of leap (366-day) years.
That needs a particular setup of the Boundary and Smooth files in lmdz.card,
and particular choices of PeriodNb in the Job.
Here the explanations, extracted from lmdz.card : 

In [BoundaryFiles] : for each year, the noleap file is copied in the run directory :

List= […]
    (${R_IN}/ATM/LIMIT/CMIP6/CM61-LR-pi-03/interpol/ELC-144x142x79_clim_limit.nc, limit.nc)

In [SmoothFiles] : 

# For leap years, copy ELC-144x142x79-leap_clim_limit.nc over ELC-144x142x79_clim_limit.nc file
List=   (${R_IN}/ATM/LIMIT/CMIP6/CM61-LR-pi-03/interpol/ELC-144x142x79-leap_clim_limit.nc, limit.nc, 3:4:),\

# As all smooth files are also copied at each job (re)submission, the "noleap" file must be re-copied over the "leap" one if the
#  (re)submitted job starts with noleap years - which is the case for exemple for the beginning of the historical period (1850,1851) 
# So, to run smoothly, the first job submission must have PeriodNb=2+multiple of 4 (ex: 18), and the following must have PeriodNb=multiple of 4 (ex: 16)
        (${R_IN}/ATM/LIMIT/CMIP6/CM61-LR-pi-03/interpol/ELC-144x142x79_clim_limit.nc, limit.nc,  1:1:2),\

# NB : There is a leap year every year whose number is perfectly divisible by four - except for years which are both divisible by 100 and not divisible by 400. 
# The second part of the rule effects century years : 1600 and 2000 are leap years, but the years 1700, 1800, and 1900, 2100 are not.
# So, the "noleap"  limit file must be also copied again over the "leap" one for years 1900, 2100.
        (${R_IN}/ATM/LIMIT/CMIP6/CM61-LR-pi-03/interpol/ELC-144x142x79_clim_limit.nc, limit.nc,  51:1:51),\
        (${R_IN}/ATM/LIMIT/CMIP6/CM61-LR-pi-03/interpol/ELC-144x142x79_clim_limit.nc, limit.nc,  251:1:251),\