What if the timestamp column or watermark is not available in the source database? How would you handle the missing field, and where would you notice the impact?

If the source database lacks a timestamp or watermark column, you still need a way to order events for incremental processing. One common strategy is to use the ingestion time as a surrogate timestamp: record the time when the row is read and store it in a new column. For streaming systems, you can generate a watermark based on the ingestion time plus a configurable delay to allow late data. Alternatively, if the data contains a natural key that can be sorted (e.g., a monotonically increasing ID), you can use that as a pseudo‑timestamp. You’ll notice the impact when downstream operators that rely on event time (like windowed aggregations) produce incorrect results or when late data is dropped. To mitigate, adjust the watermark strategy, increase the allowed lateness, or enrich the data at source with a proper timestamp before ingestion.