When ArchiveReaderWorkItem reads mailing list archives, it only parses the last 365 days worth of emails. The Mbox parser then puts them together into conversations, based on the root emails. Any email that can't be traced back to a conversation root is discarded (assumed to be sent or received out of order). All further processing of these emails, for figuring out which needs to be posted as PR comments, is then done based on conversations.

Because of this limit, if a PR was created more than a year ago, no mailing list emails will be posted as comments to that PR. This started happening when this PR was recently reopened:

https://github.com/openjdk/jdk/pull/1611

We could maybe increase the limit, but that would also increase the load on the parser, as we parse all these emails quite often. The load on the mailinglist archive wouldn't increase much as we do cache the data. We could also consider dynamically setting the limit based on the oldest current PR for the repository.