Given that this is observed in other stations too, then the TLE set isn’t good. In this case seems that the satellite is near its re-entry so the orbit changes a lot, so the latest TLE sets are getting less and less accurate from their epoch (the time that they describe more accurately the position and the other parameters of the satellite).
EDIT: I’ve just noticed that we had a gap for TLE sets for ~24h, so this also affected the issue.
I’ve generated a TLE set based on the latest good observations for CAS-10, probably is not very accurate, also there is a good chance the satellite has been already re-entered, let’s see for the next hours if we receive anything: