The documentation states clearly that "This is a retriable and idempotent operation—files in the source location that have already been loaded are skipped." What it doesn't say is that this file load history is apparently specific to the COPY INTO operation--so in practice loading even the very first batch of data with COPY INTO is necessary to have a complete loading history and avoid reloading on subsequent runs.

But, I'll happily use merge, though I haven't found correct syntax to load with parquet -- can you please share example code using MERGE INTO and parquet?