At some point they realized that for high precision work, you need to have local regulation anyway, just where the load is. These typically do not handle large currents, because each site uses their own regulators, the burden is shared. They want to use local regulation for DC accuracy, as well.
At the same time, something else happened as well: now there are linear regulator ICs with really excellent Power Supply Rejection Ratios, even at high frequencies, and they aren't even expensive. Compare to older regulators, if you try to post-filter your noisy SMPS with a LM317 or 7805, it won't do shit after a few dozen kHz, just passing it through. But some modern regulators regulate at MHz or tens of MHz ranges just fine, which is the territory where you'd otherwise need some real inductors and capacitors, which then again could have their own resonances where they are ineffective or even make the situation worse, unless you dampen them which again makes them less effective over a wider range.
But now if the linear regulator is capable of taking care up to some MHz, you are only left with really high-frequency noise which is dealt with ferrites, very small SMD caps, proper layout, shielding etc. Suddenly, an SMPS isn't so bad after all.
They need to deal with noise, anyway: it reflects back from all the other SMPS's, and this may include some very crappy ones, for example, old ones that passed less stringent regulations back then, or some illegally performing imports. If you just use a mains transformer (which has some coupling capacitance from input to output) and a 78xx style linear stage, assume it's guaranteed to be very low noise, and call it a day, you may fail due to external noise getting in through power line.
So, say, your 10x better than average SMPS design inside your low-noise product may not produce more noise than what's already getting in because of a 10x worse than average PSU in the neighbor's power outlet! Which brings us to filtration, and local regulation and local shielding.