According to https://docs.microsoft.com/en-us/azure/cosmos-db/sql-query-offset-limit recommended approach is to use token based paging, while there is also possibility to use offset+limit (aka. skip+take) approach. Because our client considers the offset+limit approach as API standard, we were asked to do a speed comparison.
Source code is available here, it runs very simple query over partitioned collection. It queries by partition key with sorting. Measured RUs in table below proves MS recommendation and shows absolute values.
|Iteration||Count||Skip+Take RUs||Token RUs||Diff RUs||Skip+Take Overhead|