Evidence for Evolution (Common Descent) has so far been indirect: (a) similarities between living organisms, (b) similarities between the living and the fossils, and (c) extrapolations of observed biological changes across a few generations. But the (a) and (b) data is often misjudged (clearly designed human creations such as transportation implements also fit the patterns), while (c) – Extrapolation from routine mutations and minor adaptations to Common Descent requires a serious Leap of Faith that goes beyond the Scientific Method. For comparison, modern Medicine is based on clinical studies (very narrow extrapolations), not on the fossil record or similarities between organisms. When this principle is ignored, we end up with casual appendectomies and much worse.
How could we actually test Evolution? Lab observation of genetic changes over a number of generations is the ideal experiment, and this has been done many times, but not as a pass-fail test of the Evolution hypothesis. To confirm Evolution and separate it from mere adaptation or genetic drift, a Baseline and a Minimum Threshold have to be established before the actual experiment is conducted. The ideal organism to be tested is very small and is rapidly reproducing. The Baseline would be a fully documented single organism from which all subsequent generations descend while the Threshold for success would be a descendent that under no circumstance would be classified same as the Baseline organism. This may be contentious as biological classifications such as Population, Strain, Species, Genus, etc. are all subjective.
How long should the experiment last? If E.coli is the test organism (10^9 organisms/ml and 20 min/generation), it takes only 1 (one) liter of E.coli saturated solution and 1 (one) hour to obtain as many individuals as were needed for human evolution from ape (5 Mil years, 10 Mil/generation and 20 yrs. /generation). E.coli are not humans indeed, but neither are the laboratory conditions same as natural ones as the main advantage of the laboratory is to accelerate testing in a controlled environment by introducing specific inputs at a much faster rate than in nature, and by monitoring their effects to adjust these inputs.
Other test considerations include: What exactly cause the transformations observed? Do the new strains have to trade off new features for old ones or are they overall better off? What are the limits of transformation? What happens with the new population if the stimulus is removed? Do the various strains turn into a homogenous population? Does that population retain its transformations or does it revert to the Baseline?
Genetically Modified Organisms (GMOs) may also prove Common Descent despite these being primarily examples of Intelligent Design. Venter’s Minimal Bacterial Genome in particular almost meets this Threshold if the new organisms were not just degraded versions of the Baseline. To date, other GMO experiments have obtained enhanced versions of the original, but not new organisms that meet the Common Descent Threshold as defined here.
Can we use any natural experiments? Of course. We can look at ancient remains that preserved organic matter including DNA. Based on the theory of evolution, the further back in time, the more differences between the ancient and current organism we would expect to note. And, given the shorter life cycle of microorganisms, we would expect to see many more differences in these organisms than in larger, longer-lived ones. The ancient organisms would also be inferior in survivability to current ones based on the evolutionary arms race hypothesis.
Observations incorrectly presented as proof of Evolution include: Domestication, Adaptations (the beak, the moth, camouflage, etc.), Antibiotic Resistance, Cancer, Immunity Response, Horizontal Gene Transfer, Hybrids, Dimorphism, Polyploids, and Mathematical Models. All these are fine examples of adaptation, but none meets the Common Descent Threshold as defined here. In addition, Cancer is suicide by death of the host and Models are biased human creations that seldom match reality.
Often cited Failed and Irrelevant Experiments: Richard Lenski at MSU did not obtain anything other than his baseline E.coli, Barry Hall at U. Rochester and Paul Rainey at Oxford only confirmed adaptation, Miller Urey did not demonstrate abiogenesis without which Common Descent is a nonstarter.
In conclusion, verifying Common Descent in a laboratory experiment is feasible. However, all previous attempts failed and/or were improperly reported as partial successes.
Con: What about beneficial traits spreading throughout the population?
Pro: Indeed, a verifiable new trait that proves beneficial enough to spread unaided throughout the population would be a first (although insufficient) step towards proving evolution.