Request for an advisory on best practice and possible performance pitfalls

Scenario:

Given a table that continually has rows added, we want to purge entries fulfilling a certain criteria. However, we want to persist entries before purging and we want to ensure that only persisted entries are purged.

Solution (so far):

We do this as a scheduled Job, that executes 3 steps implemented as separate procedures:

• Step 1:
  Copy entries from source table that meets the criteria to a structurally identical table (henceforth: the buffer table). The buffer table is only accessed by the Job.
• Step 2:
  Save rows from the buffer table to the file system.
• Step 3:
  Delete rows from the source that are present in the buffer table. Rows are deleted in chunks in order to avoid long locking.