The segment shows sessions where the number of sessions to transaction is more than one (so 'returning' visitors making a purchase, in theory) and where the days to transaction is zero (the sessions were actually on the same day). Many of these orders were probably part of single 'real life' visit which was split by the standard sessionisation schema.
The segment works well on the 'Sessions to Transaction' tab of the 'Time to Purchase' report.
Behind the dramatic title is a rant I've made so very, very often in workshops, trainings, and consultation work.
Have you ever actually stopped to consider just what metrics you are optimizing against when using a web analytics platform? Google Analytics, for example, stitches the raw hit-level data coming in from the website into Sessions and Users. The latter has some grounding in the real world (all hits shared by a single clientId), but the former is completely arbitrary, artificial, and irrelevant.
Never mind the vague description of what constitutes a session in Google Analytics, because it rarely has anything to do with the thing we're really trying to convert against: a user with some specific intent. Intent is tricky, since it can span across many "Sessions", devices, days, weeks, and even websites. But that's what we should be interested in.
Conversion Rate, for example, is an inherently flawed metric, as it's bound to the concept of Sessions. Change the definition of a session even a little bit, by e.g. increasing or decreasing session timeout, and Conversion Rate will change.
I wrote this article to vent, but I have found that many seem to forget what I consider the basic tenet of data collection and processing: the numbers you see in reports make sense only if you understand and accept the underlying schema.
So this is more a call for critical thinking than a request to change how these tools work (though I do rant a bit about this as well).