So there's different source formats currently exposed, depending on the configuration of the MongoDB connector:

• When using "oplog" capture mode, "h" or "stxnid", "which represents the lsid and txnNumber fields from the oplog event"
• When using "change streams" capture mode, "lsid", which apparently resembles "stxnid", but in a different format

Let's see whether we can stick to the current format also when using change streams, or whether we should start exposing one opaque "sequence" value, as is the case for the Postgres connector already.