Yes, you're correct, the trend is strongly portable devices with crappy earbuds, but trust me, I've been on the other side of this...there's not much consideration at all to the listener's environment, not in music or radio. The problem is there's no acceptable average listener environment that you can process for that doesn't mess with other environments. The music industry is very centered on their own personal stamp on the creative process, so producers simply ignore facts, and stamp away. The same is true for radio broadcasters and cars. There actually is data on noise levels in cars, but broadcasters really don't care.
In both cases, the rule is, I want my stuff louder than the other guy, or at least as loud, but certainly not quieter! It's a guy thing, major male ego trip, pissing contest, whatever. If you were responsible in part for the financial success of your product, and if you were in a bar and your record came on, but it was suddenly harder to hear than the other records, how would you react? If you were listening to a competing radio station, and hit the button for yours, but it didn't sound as loud...same reaction...."What the!!!!!" And, you'd want to turn it up.
Now, the issue of processing for the listening environment. Yes, noisy environments are a problem, but much more so for some genres than others. Anything popular and contemporary has little dynamic range, even if not processed for loudness. It's a problem with a 100dB noise environment, but nothing like classical music, which is unlistenable that way. But, the processing could be, and should be in the player. How hard would it be to put a little mic on the player...most already have them...sample the noise, and adjust a dynamics processor running in on-board DSP to take care of it? This should be easy, and, if anybody at RockBox is listening, stuck into the next build.
In the days when HD Radio (man I hate that name!) was in development, I suggested the bit stream include a "rosetta stone" data block that represented a reference for post-processing. That data block would be used, along with an environmental noise sample, to adjust dynamic post-processing in receivers. It would represent a profile of the processing that had already been done, and thus be the "rosetta stone" to un-do that processing, or do more along the same lines. Basic receivers could ignore the data, but high-end ones would also have user controls, so if someone wanted a non-processed version, he could pretty much have it except for any in-studio processing on mics, or slow compression for sloppy operators. Of course, there was no interest, and at the time I worked for a company supporting HD radio development in another direction, so I was kind of gagged anyway. But the lack of interest was mainly because broadcasters LOVE their loudness wars! At least, when they win. It's a HUGE thing.
One more anecdote so you see how massive loudness is as a motivation. I interviewed for a chief engineer's job at an FM station once, the GM told me within the first two minutes that it was his goal to be the loudest station on the dial, and he would spend whatever it took to get there and stay there. He'd already spend $20K (that would be in late 1970s dollars) on the latest/greatest/baddest processing package, and wasn't satisfied yet. That was decades ago, and you'd think the introduction of HD Radio would have tempered the war. Nope, not until you temper the egos first.
I'm sorry, I can't help being cynical about this. I just don't ever see it improving unless loudness processing has a quantifiable negative impact on music sales, and right now there are enough other negative factors in music sales that it would get lost in the noise...if you'll pardon me.