"Back in the day" it was common to have a "racing bike", or at least "racing wheels", and a set-up more generally used for training. The idea was the training bike would take more wear & tear, while the racing bike was kept in more refined condition for racing where small efficiency gains were truly important. Examples of tradeoffs include:
- cost: given a fixed amount of wear & tear or damage risk, it's better to use a cheap part than an expensive one, since the cheap one is easier to replace. Examples of this include wheels, cassettes, and chains: these parts wear out (wheels from rim brakes, drivetrain from normal use) and if the performance gains of expensive stuff aren't important in a particular ride, it makes sense to go with the cheaper parts for these.
- reliability: lighter parts tend to be less reliable, with less wear life, so even at the same cost (which they're not) it makes sense to stick with more durable stuff if the truly marginal advantages of grams saved are unimportant. Thicker innertubes or tires are another example of trading reliability for performance.
- condition: Unworn, clean parts tend to be more efficient than worn, dirty parts. Consider chains. Chains stretch and wear, and as they do, drivetrain efficiency suffers. Losing 0.5% of your power because of worn equipment may be irrelevant on a solo training ride, but may be critical in a challenging road race or time trial. It makes sense to get more use out of parts in a training context than in a racing one.
In addition to these quantitative differences in economic optimization there's an additional issue, which is psychological. Riding a training bike for awhile, then getting on a racing bike, feels like a performance boost: the racing bike feels faster, whether it actually is or not. This little dose of confidence can be valuable in challenging performance-critical situations.
Training tends to be divided between group rides and solo, with the group rides fairly often race simulations where riders would compete with each other. The importance of these competitions was subject to debate: for example, if one bike had a 10-second penalty climbing Old La Honda, but the result was you were waiting 10 seconds less for the the slowest rider to arrive, was that undesirable? There's a strong argument that, to the contrary, it's desirable: 10 seconds of extra climbing is 10 seconds extra training which means 10 seconds more fitness, assuming you still finish the ride. But for the last rider, it might be undesirable, since it may mean the group leaves before you arrive, or you keep them waiting longer, or you're denied 10 seconds of much-needed recovery at the top.
So "back in the day" riders would time themselves up climbs, or keep track of how they finished in town line sprints. Typically these results were completely unimportant, but other times, for example getting a PR on a climb, results had a more lasting impact. But generally it didn't make much sense to use top-tier equipment on training rides.
Then came an era where among a certain segment of the riding population started using power meters. With analysis of power data, actual speed took a back seat. Focus was shifted to how many watts/kg of body mass could be sustained for a certain time, or if you were sloppy on the body mass part, just watts. Heavier bike? No problem. Inefficient drivetrain? No problem if you measured power upstream of the drivetrain (for example, at the crank as opposed to the hub). Rolling resistance because of thicker inner tubes or tires? No problem.
But now we're in the Strava era. Every segment of road is measured and documented. For some, actual events are more opportunities to score good Strava times than to generate event results few will see or care about. Unlike the power era, time matters again. And unlike the pre-power era, the time doesn't matter just on your favorite climb or sprint (for example, Old La Honda times) but on virtually every ripple in the road. There's added incentive for riders to bring out the high-performance hardware, for example carbon-rim wheels and lightweight cassettes, on relatively normal rides, even commutes.
But I cling to the tradition that a training bike should be about training and not about shaving seconds. On MDR, which I just completed, I brought my tried-and-true steel Ritchey Breakaway. The carbon Fuji SL/1 was safe & sound on the bike rack, carbon tubulars and associated brake pads in place, saving it in case I decided to do the Pescadero Road Race, this coming weekend. When I struggled to stay with the group on a climb, I was in theory working slightly harder than I would have otherwise, and in the end that has the potential to make me stronger.