
There are occasions when one wants to use ISTART=8 to start the model
up from a previous run's rsf file but avoid some initializations, so
that the model acts more like it's a continuation of a run (like ISTART=9.)
It is not always easy to remember how to do this, so below let's compile
a list of things the user has to do to get the prognostic variables
at least to be the same as they would be with an ISTART=9 setup. Much of
this is taken from Reto Ruedy's emails and commit notes:

    NINT Case
    =========

___ set parameter: keep_params=1

___ set parameter: init_topog_related=0

___ set parameter: init_flake=0

___ set parameter: niprnt=1 [Only needed if you want to print out the
    initial conditions and create the PARTIAL acc file.]

___ The radiation must be called at the same times as in the original run.
    This should be automatic if nrad=1; with nrad=5, it is automatic for a
    365-day year if you start at the same calendar day as the original run.

Reto notes that, even so: "only the prognostic variables have the same
values; some diagnostics may be slightly different (unless all diagnostics
were called at that starting point also in the original run)." He thinks that
diagnostics could be made to agree by setting their "printing" frequencies
all to 1 timstep, but that would slow the computation down with little gain.

    Tracer Run Case
    ===============

___ all of the settings above for NINT, of course, because most tracers
    have some interaction with the background weather.

___ set parameter: allowSomeChemReinit=0 [Only relevant if you have the Shindell
    chemistry compiled; tells the model to skip some initializations. Namely:
    setting of H2O and Q in the stratosphere consistent with CH4, applying
    only a portion of the chemistry change in the first 30 timesteps of a run,
    initializing some non-tracer species and intra-family proportions, reading
    default albedo-related variable from file. On E2.x branches, that last bit
    also requires keep_params=1.]

___ set parameter: initial_GHG_setup=0 [Only relevant on E2.x branches and
    earlier.]

___ set parameter list: itime_tr0. For example: itime_tr0=-1,-1,... out to ntm,
    to prevent tracers from reinitializing when itime = itime_tr0(n) or from
    skipping tracer code because it thinks your tracer has not reached it's
    turn on time yet. The -1 suggestion helps because itime should never be negative
    so itime should always be > itime_tr0. Obviously, if you use a tracer that
    should in fact not be turned on yet in your run or depends on itime_tr0 in a
    more complicated way (bomb isotopes?) then set this as needed. For OMA tracers,
    I think the only dependence is the GLT chemistry diagnostic tracer, and if you
    set this incorrectly and don't care about it, the change in values will not
    affect any other model variables. The GLT issue only matters on E2.x branches
    and earlier.]
    Note: on E3 branches or later one can set e.g.:
      default_itime_tr0=-1  ! default
      GLT_itime_tr0=12345   ! exception
    instead of a long comma separated string of numbers.

As noted above for NINT, diagnostics may still differ vs. an ISTART=9 start. Also,
the 'Water' tracer still currently exhibits differences, as does many tracers-in-
surface-reservior variables. I.e. something about the tracer interaction with the
surface hydrology is still affected by the ISTART choice, and while this doesn't
seem to affect atmospheric tracer values as a whole, it seems to matter for 'Water'
tracer.

