Setting time as an unlimited dimension in netCDF

[ctroupin]

Shouldn't we define the time in the netCDF file storing the climatology as a record dimension?

A dimension length is an arbitrary positive integer, except that only one dimension in a classic netCDF dataset can have the length UNLIMITED. In a netCDF-4 dataset, any number of unlimited dimensions can be used.
Such a dimension is called the unlimited dimension or the record dimension. A variable with an unlimited dimension can grow to any length along that dimension. The unlimited dimension index is like a record number in conventional record-oriented files.

Doing so, we don't have to edit the files with nco to merge new files with previous products.
Also maybe the dimension observations (used for obslon, obslat etc) could be unlimited. To discuss.