Hmm . . . I've seen a similar sort of plot, in the past.
If I am remembering correctly, it was
NewtronStar's circuit that showed similar oscillations. We never did discover what initiated that behaviour but found that by powering down the modem (or combined modem/router) overnight and disconnecting the patch lead from the device to the SSFP, just the once, cured it. (I described the process as "giving the line-port card sufficient time to relax".)