Liquibase and JDBC driver cancellation: should interrupted changeSets clear the interrupt flag or propagate it?

I am integrating Liquibase update execution into a service that manages its own worker threads and cancellation lifecycle. The integration boundary in question is between Liquibase's changeSet executor and the underlying JDBC driver: when a long-running changeSet is cancelled via Thread.interrupt, Liquibase attempts Statement.cancel() if the driver supports it, and otherwise lets the thread exit with InterruptedException.

The uncertainty is driver-dependent behavior and interrupt-flag ownership. PostgreSQL implements cancel() via a CancelRequest message, while MySQL Connector/J may ignore it or throw UnsupportedOperationException, so the same cancellation request produces different latency and exception types per driver. Separately, query timeouts set via Statement.setQueryTimeout surface as driver-specific SQLExceptions that Liquibase treats as changeSet failures, but there is no unified Liquibase configuration to choose between interrupt-based cancellation and timeout-based termination. Assume Liquibase 4.x with recent PostgreSQL and MySQL drivers.

My questions:

1. After Liquibase catches InterruptedException during a cancelled changeSet, is the thread's interrupted status preserved for the calling framework, or cleared? Is this documented or driver-dependent?
2. Is there a supported way to enforce a portable statement timeout across drivers, or is a custom Connection wrapper the only option?
3. After an interrupt-driven abort, is the JDBC connection guaranteed reusable, or should the pool discard it?