I'm not entirely sure I understand your questions but I will state the following:
When a node runs the data associated with an output pin's records are stored in an application temporary data file. The size of this file is not affected by how many records are configured to be displayed in the data viewer (since all of the records are always stored).
The ls.lae.dataviewer.pageSize property only affects the maximum number of records that will be displayed by the Data Viewer. However, the data to be displayed must be transferred from the server-side to the client browser. So, if the property value is increased then more data will be transferred when the output pin has more records than the current limit (1000) - requiring additional bandwidth and processing on both the server-side and client (browser). Compression of HTTP responses from the web app was introduced in the v.3.6.0 release, which will partially mitigate the extra bandwith required for transferring the additional records and also increases the responsiveness of UI in general. Increasing the property value will increase the data transferred to all clients so this may increase delays in displaying data, especially for clients on slow internet links when the data set is very wide. The system maximum of 20,000 was set as a protective control to limit the overall impact on the operation of an Analyze Server intance.
There is no per-node control to determine the maximum number of records that are displayed by the Data Viewer. A Head (or Tail) node can be used to extract a defined maximum number of records from a data set of an upsteam node.
The Data Viewer allows (subject to permissions) a user to download the entire data set for an output pin. This data is downloaded by the client browser as a zipped file.