This has one drawback: after flushing log data to its underlying output,
a new NXLogStamp object MUST be written to the NXLog object in order to
create a new entry within the stack.

This can be changed if necessary. For now I'd like to keep it as-is.