seqId is used to find which key's value is the most recent in the file. However, if the index is loaded from the disk, it is never set and the new entries are stored from seqId=1.

Reads/Writes aren't affected but iteration relies on this to discard old values. It has the side effect of (possibly) returning multiple values for the same key.