Feature Request: Indexable writes to netCDF format files on disk

This is a feature request for indexable writes to a netCDF format file.

The documents mention that data variables (my interpretation is: the entire multi-dimensional array represented as a data-variable entity) can be appended-to or overwritten when using ds.to_netcdf(<filepath>, mode='a'). However, this requires first reading the dataset into memory, then modifying some-or-other data element at said index in a multi-dimensional array, and then replacing the entire data-variable on disk.

It would be incredibly useful to be able to write a single data value to a single array location without first loading and then completely overwriting the complete data-variable.

There may be ways to achieve a similar result by using the underlying netCDF4 module, e.g. per this stack overflow post. But forwards-compatibility and unintended side-effects put a damper on such approaches.

Start by tracing the ds.to_netcdf(, mode='a') entry point and compare its behavior with the underlying netCDF4 indexed-write approach referenced in the issue. The desired outcome is writing one value at one array index without loading and replacing the complete data variable; no specific files or tests are named.