are checkpoints guaranteed to be resumable across version upgrades? and if so, when the job takes its next checkpoint after a version upgrade, will it wind up being a continuation of the incremental checkpointing or will it behave as if it was restored from a savepoint and it’ll “look like” a full checkpoint?

See https://nightlies.apache.org/flink/flink-docs-release-1.16/docs/ops/state/checkpoints_vs_savepoints/

aha, that’s precisely what i was looking for, thank you! a version upgrade for an aligned-checkpointed job should be safe and continuing incrementally checkpointing, then 👍