You mentioned that this only happen to a specific source, perhaps there are problems with that source (e.g. unreliable connection, loaded server)?
You can attach here or email the support team (firstname.lastname@example.org) the log file containing the bogus chunk and we’ll check what inbound stream could be causing it and confirm it with you.
As for the segments not being found as reported by nginx, perhaps you have configured hls to be a rolling type and there are cases that this segment has already been deleted because it has reached the number of rolling segments. Also, the nginx server could be caching the manifest file for too long, giving clients an older list of segments that are no longer existing.
Hope this helps.