Run simulation length seems out of sync between EAMxx and CIME

There appears to be a discrepancy between the length of simulation run reported in the e3sm_timing file vs. the actual length of the run, when running with EAMxx. This was first noticed when looking at PACE results, which always reported a run length that was 1-day short.

It may be related to the fact that EAMxx has 0-based indices rather than 1-based. But it isn't clear if ATM reports nsteps back to CIME, or CIME keeps track some other way.

Making this issue to have a central location for discussion.

Start with the e3sm_timing file and trace how EAMxx reports nsteps to CIME, comparing that value with the actual run length observed in the PACE results. Done means the source of the one-day discrepancy is identified and the reported simulation length matches the actual run.

Written by the indexing model from the issue text.