1) 400 yrs seems enough if you start cndv from a spun up CN state. The compset that you used (I_1850_CN) should point to a spun up initial file. You may need to use the interpinic tool to get that initial file to work with your cndv run. The clm user's guide has documentation on interpinic.2) The initial file from the offline case will either work or it won't in your coupled case. It won't work if variables have different vector lenths in the two simulations. Unfortunately interpinic does not map initial data from one cndv simulation to another. If you run into this problem, then you may need to do the cndv spin up in coupled mode starting from a spun up CN state.Sam