It is possible and sometimes expected that MySQL jobs in Airbyte may force a full refresh after an upgrade, depending on the nature of the upgrade and the connector versions involved. For example, when upgrading certain source connectors (like Snowflake), the documentation explicitly states that incremental syncs will trigger a one-time full refresh after the upgrade because the old connection state is not compatible with the new connector. After this initial full refresh, incremental syncs will resume as expected. While this example is for Snowflake, similar behavior can occur with other connectors, including MySQL, especially if there are breaking changes or state incompatibilities introduced in the upgrade process [Upgrading to 1.0.0 - Snowflake](
https://docs.airbyte.com/integrations/sources/snowflake-migrations#upgrading-to-100). Additionally, in the context of MySQL CDC (Change Data Capture), if the binlogs required for incremental syncs are no longer present (for example, due to log retention settings or a long pause in syncing), Airbyte will trigger a full refresh sync from scratch. This can happen after upgrades if the state is lost or incompatible, or if the binlogs have expired during the upgrade window [MySQL Troubleshooting](
https://docs.airbyte.com/integrations/sources/mysql/mysql-troubleshooting#under-cdc-incremental-mode-there-are-still-full-refresh-syncs). There are also user reports indicating that after upgrading Airbyte or its connectors, connections may trigger a full refresh or even clear data unexpectedly, especially if there are schema changes or state management issues during the upgrade process. This has been observed in both MySQL and other connectors, and sometimes relates to how Airbyte manages connection state and schema changes across versions [GitHub Issue: Data loss when reactivating a connection](
https://github.com/airbytehq/airbyte/issues/55770).
In summary: Yes, it can be expected for MySQL jobs to force a full refresh after an Airbyte upgrade, particularly if there are breaking changes, state incompatibilities, or binlog retention issues. If this behavior is unexpected or causing issues, review the upgrade and migration documentation for your specific connector version, and check your binlog retention settings to minimize the risk of unnecessary full refreshes.