Monitoring
What happens when a data source changes?
Distinguish real changes in the underlying data from collection failures and missing values.
Separate three different signals
A value changing, a source removing a record and a collector failing to read a record require different responses. Compare the latest observation with prior runs and retain the source and observation time for each record.
Investigate unexpected drops
Check whether the source is available and whether its layout, API response or field definitions changed. Compare record counts and required-field coverage with recent runs to identify changes worth investigating.
Make exceptions visible
Agree which changes warrant an alert and which can wait for the next delivery. Label stale or partial data explicitly so downstream users can decide whether to use it.