I had filed case 00739580 on 09/19/2018 (text below) and concluded that it looks like a bug. The system is tracking the ids of the subscribers instead of the emails so if one sends an email to subscriber id 100 with email Email1 then changes the email of subscriber id 100 to Email2, the bounce report shows that Email2 bounced while in reality, Email1 is what bounced.

We have encountered what seems to be an anomaly.
An email "send" took place on 9/10 by the name of SUNY-30. A subscriber by the name of Skype Lagombra, EmailActivityID (our unique id) SA0147437, email skye.lagombra@bronxdalehs.org was sent an email in that 9/10 "send".
on 9/13 roughly we uploaded a file with the option insert/update that changed the email of the subscriber to skye.lagombra@gmail.com. Another email titled "Thank You..." was sent.
on 9/18 we pulled some data (export) on the 9/10 SUNY-30 "send" to measure the results of that "send" and found the anomaly. The Sent file for the 9/10 "send" shows that an email was sent to skye.lagombra@bronxdalehs.org (which is correct) but the hard bounce file shows a hard bounce to skye.lagombra@gmail.com which is not correct. The email skye.lagombra@gmail.com didn't exist in the 9/10 "send" so it shouldn't factor into that "send" results.

The expected result is that the export that we perform, to study the results of the email send, should contain the results of that particular send unpolluted of future changes to the list so it should show the results of a send to skye.lagombra@bronxdalehs.org.
It seems as if you are capturing a snapshot of what was sent (i.e. the actual email, name, etc.) but you're capturing the subscriber IDs of what bounced, just guessing.

We would appreciate any insight you can provide if this is working as designed.

Thanks,
Eddy