We have a requirement to demonstrate a Reliability of R90C60 but due to resources limitations we can test only 2 - 6 machines (the lifetime runs in hundreds of hours and it is expensive to re start the test after a component failure). In this scenarios what kinds of life time test you will suggest and what to report to customers?
This also brings a question in my mind that we sometime struggle to answer when discussing with marketing. How to define the 'Lifetime' of a product because you can always repair or replace a component and increase the lifetime of the product. For e.g in a car you can replace the complete engine or rework the chassis and the lifetime of the car will increase accordingly.
In my experience, "lifetime testing" usually means how long with the system keep running until something causes it to fail, to stop doing its function. Now perhaps you could have a "minor failure" that would not affect the operability of the system, and I would keep the machine / system running. But I would say that once the system stops functioning, and you have to do something to repair it (or replace a component) then that was "end of life failure".
Again, depends on what the customer is asking for in the specification. For example, in the "real world" you do preventive/predictive maintenance on you automobile. By the fact you change the oil every so often, you extend the life of the automobile. In the "real world" I am also looking at, through Corrective Maintenance and Preventive Maintenance programs, what is the expected life cycle dollar cost to maintain the machine, and what can I expect for an availability rate (Mean Time Between Failure and Mean Time to Restore).
Without knowing the details of what you are testing, and more important, WHY are you testing it, hard to give firm advice.