PIT was implemented in the first iteration, but didn't work in production due to differences between the local ElasticSearch container and the AWS ElasticSearch service, which had removed the feature. As we are now on OpenSearch and have access to PIT again, it might be worth switching to using it.
Needs investigation of the impact on the cluster, and if/how we can share a PIT between workers (I think that creating individual PITs for each job might be worse (and wouldn't fix the "records moving buckets during generation), but maybe not?).
PIT was implemented in the first iteration, but didn't work in production due to differences between the local ElasticSearch container and the AWS ElasticSearch service, which had removed the feature. As we are now on OpenSearch and have access to PIT again, it might be worth switching to using it.
Needs investigation of the impact on the cluster, and if/how we can share a PIT between workers (I think that creating individual PITs for each job might be worse (and wouldn't fix the "records moving buckets during generation), but maybe not?).