Depends on what you're optimizing for. If you're considering solar purely from the angle of replacing carbon-thermal generation on the grid then sure, large arrays are more efficient, offering more room for amortization of fixed costs, more optimization for solar gain, and more room for optimal hybrid usage like this example.
However, buildings need roofs anyway, and will in turn get sun exposure. Since solar tech like tiles can take the place of traditional roofing, there are some double gains to be had there in that they're both doing the job of protection from the elements and taking otherwise mostly wasted energy and doing some work with it. Depending on how one gets into the weeds on aesthetics (like if they wanted nicer tiling anyway) the marginal extra capex of tiling may well be worth it as costs come down further. Building-solar also can help provide resiliency to grid damage, which by definition grid feeding cannot. For people in areas where they'd otherwise be running generators anyway, solar/res-wind+battery (and as BEVs take over near everyone will have an extremely sizable slab or three of battery around much of the time) can be compelling. Still more upfront, but maintenance-free for a decade or more and constantly providing some ROI (and effectively constant verification everything is working), whereas hydrocarbon generators require regular maintenance/testing which cost money and generate zero return otherwise, they just depreciate. And local solar/wind/utility resources are going to affect the time horizons for all this.
So basically there are a ton of new variables and enormously more scalability up and down the spectrum for renewables and batteries. Doing the math is in turn going to be very individual, but it will often still make sense to do both.
Also:
>why don't we do this more often?
I mean, we're still in a pretty steep part of an S-curve here. It's just plain early days. People are still experimenting with stuff like this and learning what works. Unit costs are dropping, which in turn changes what projects make sense which in turn changes demand and thus unit costs. Grids are adapting and getting smarter. Both storage and opportunistic demand are doing the same in parallel in a variety of ways. Stories like this where someone tries some new stuff and it works out well will make others perk up and take notice. There will also be things tried that don't work out. Going to be a wild decade.