I don’t know for fact that Polaris does it this way, but would bet money that they do... “Test” units usually go through different iterations. This would be a general progression. Different manufacturers and even different projects will vary slightly, but the general theory is likely the same. There are often stages in between what I’m laying out below, but this would cover the higher level steps.
Step 1: skunk works prototype, “mule”. Kind of a proof of concept. Hand built, using non-production procedures and components. A lot of times the “testing” done here is more around design concepts and not so much around repair frequency tracking.
Step 2: full prototype. This is where you’d have maybe 70-80% final product components and some level of production procedures, but not many. These are the units that would go through the majority of the testing and failure analysis. Many engineering changes and component changes based on what they learn. The units are updated periodically with new designs and testing continues
Step 3: pilot units. This is as much a test of manufacturing process as it is anything else. The product is 95% or more “final” product. These units will have some specific test requirements and used for auditing numerous aspects of the project. Everything is built in the production environment. Very few changes to the product can be made at this point. If there are, step 2 was a failure and there will be a delay in the project for additional test needs.
To say engineers working on these projects don’t have real world experience is complete ignorance. Yes, some (many) design engineers fit that label, but design engineering is only one piece. Test engineers are often mechanics, engineers, riders, users, etc... they don’t sit behind a computer screen in CAD all day, but they can... there are also technical service personnel that are very involved in these projects that bring data from past experience, serviceability, warranty issues, etc.
There is a lot of hyperbole in this thread. What Dan proposes in all these situations is what we consider a “10X” fix. Basically, spend any amount of money, over engineer it, and push out a fix that is almost certain to never fail (10x stronger/better than it needs to be). IF Polaris starts experiencing a 20+% failure rate throughout the entire vehicle park, that might be something they consider. But I doubt that happens or the project wouldn’t have been released. If they see failures, they are going to go into failure analysis and fix the root cause of the issue...not just do a 10X fix because that changes down steam variables.