Correct answer is A. This is classic A/B testing. Since you already have a new version, built into an image, all you need do is to use the load balancer to split traffic going to old version and new version. See: https://cloud.google.com/load-balancing/docs/l7-internal/traffic-management#traffic_actions_weight-based_traffic_splitting. Note that global load balancers can route to serverless services. https://cloud.google.com/load-balancing/docs/https/setting-up-https-serverless

Besides Cloud Run alone doesn't provide Feature Flag solution that requires Firebase Remote Config feature. For an A/B test you have to fix what users are using what version. Option A doesn't not provide a complete solution for something that is crucial. So, besides both seems incomplete, if that request was made to me, I'd implement as a Feature Flag using Firebase Remote Config

Considering the importance of traffic analysis and the need for precise control over traffic distribution for a global ecommerce web application, Option A is likely the better choice. This option allows for detailed monitoring and analysis of user interactions with different color schemes, offering clear insights into which version performs better in terms of sales. The use of an external HTTP(S) load balancer for traffic routing provides a more controlled environment for conducting such a study.

A is correct.

A is right. D is specifying 50% of the users which is not correct. In really the traffic split is 80-20 or 75-25 ratio. This is a specialized version of canary depolyments.

This is the best way to test the new color scheme on live production traffic. By enabling a feature flag, you can display the new color scheme to a subset of users while keeping the old color scheme for the rest of the users. This will allow you to compare sales between the two groups of users and determine whether the new color scheme has a statistically significant impact on sales.