Read-only Grid Paging Algorithm

During record sync the data will be queried from database and stored in Appian record type. This is like a virtual data layer when data is kept for optimised data storage and retrieval within Appian's data architecture. 

In a grid when you page, the first batch shows first 10 rows of that record type, on click of next, the next 10 are retrieved from record type and shown. 

You are thinking in correct direction here. No database transaction happens when grid page is modified rather user is working with Appian data fabric layer to see the next batch of data which is the optimised way to working with data in Appian. Hope it clarifies

When you click Next on a sync-enabled record-backed grid, Record grid makes a database query to its synced database (Appian Data Server) to fetch the next 10 rows. No call is made to your source on pagination.

https://docs.appian.com/suite/help/26.5/about-data-sync.html#optimized-data-access

When a user clicks Next in a grid backed by a sync-enabled Record Type, Appian sends a new query with updated paging information (startIndex and batchSize). The data is retrieved from the Appian Data Server, which stores the synchronized record data. Appian does not load all records into browser memory; instead, each page request retrieves only the required subset of records from Data Server, making pagination efficient and scalable.