Ipsos did a recontact survey of its June 20th poll in the battleground states.

It found a 4 point shift to Harris with this group.

It was a recontact survey, so the same respondents were sampled both times. Thus sampling error is minimized.

It is nice to have confirmation that this is a true shift for Harris and not just differential partisan non-response or sampling error.

Also this was pre-debate, so if anything it probably underestimates the shift that resulted from Biden withdrawing.