Stepper: Figure out why the wide discrepancy on signup start referrer between start and stepper during the experiment

[MicBosi]

Slack conversation: p1732718523042509-slack-C073776NJ66

It appears that stepper is getting the majority of the traffic from the LOHP without locale suffix, while control is getting most of its traffic from the LOHP with the locale suffix.

Control also gets practically all the other referring traffic (google, bing, random sites etc.) and stepper appears to be getting none of it.

However, if we look at the distribution of languages (based on the browser lang reported by Tracks) we appear to see languages roughly evenly distributed between the two variations.

We would like to understand the mechanics of how this happens so that we can explain whether there is a qualitative difference between the traffic sent to stepper vs the one sent to control that can explain the SRM issue at the signup start step.

[MicBosi]

I think we might have found the cause of the issue being the fact that all Stepper traffic in the experiment is redirected in JS/Calypso using Referrer Policy: origin which truncates the referrer to just be the domain, in this case wordpress.com

See full explanation in Slack: p1733786423894269-slack-C073776NJ66

[MicBosi]

@escapemanuele I'm closing this issue as per Slack conversation mentioned above I think we have a satisfactory explanation.