For state-enabled base objects only, the tokenize process skips records that are in the DELETED state.
These records can be tokenized through the Tokenize API, but will be ignored in batch processing. PENDING records can be matched on a per-base object basis by setting the MATCH_PENDING_IND (default off).