That's what troubled me too, I asked the same on discord couple of weeks ago and got this answer

There is a bug in Keychain where some transactions are broadcasted twice. I guess this is something related to that. Even though two custom jsons were processed at the same time, Hive Engine would have accepted only one.

Then I headed down to Hive Engine explorer and saw this

First image ( accepted stake )

Second image ( rejected stake )

So the payload for both the custom_json was the same but the 'logs' was different . So if you want reliable 2nd layer token JSON data , according to me , hiveengine API is where to look at.

HiveSQL stores 'payload' in its table but what we need here is actually 'logs' . So right now I am extracting all the data from December 1st and storing it in JSON format ( although I prefer CSV , it will become difficult for me to clean it later) .

Did any of the above make sense?

!WINE

Posted Using LeoFinance ^Beta