LOTW has some serious problems. I'm not suggesting it is in danger of going out of business. But it is not obviously a very good business.
1. Like a lot of web services, all the payment is back-ended. They offer a free service until you cash in. There's a lot of stuff on the internet that works like that. A close analog would be in-app purchases on an iPhone for an app that is free until and unless you decide on some purchases. Maybe you pay to get rid of ads, for instance. Or, maybe you pay to get a cool "skin" for your iPhone game character. Yes, people do that.
2. The economics for LOTW, then, is to get enough people to "convert" from unpaid to paid. In our parlance, to actually apply for an award. The "conversion problem" actually a studied problem for people who do things like invent iPhone apps that are free-with-something-to-buy-later. However, most people do seem to want DXCC and similar awards, so this is the least of their problems, fortunately, particularly with the tie to the CQ WAZ stuff and even the WPX stuff.
3. One problem that maybe sets LOTW apart a little bit from, say, some iPhone gaming app is that it has some dis-economies of scale. A substantial fraction of uploaded QSOs, particularly those that go more than 120 days without a match never get one. Those are still potential money-makers, but the majority never do. Worse, there are a lot of matches that never get turned in. I got my 20m SSB QSO with Japan confirmed in the paper days. Most newly minted DXers get theirs in LOTW about the first week they take up DXing. The thousands of other matches to JAs (e.g. in contests) are all dead bits whether they match or not. And, since those dead bits reside in some sort of curated data base the cost per QSO (dead or alive) is much greater than the raw cost of storage. Maintenance of data bases costs real, skilled people to manage. It isn't just shoving the next 2 Terabyte Drive into some slot.
4. The pile of dead bits grows every day. So do the moneymakers. But as someone's career in DXing matures, pretty much everything is dead bits. Since I don't contest much I "only" work a couple thousand Qs a year. About half confirm in LOTW. But, of those, since I am north of 2400 on the DXCC Challenge, maybe one hundred at most end up generating LOTW any money. So, the longer I DX, the lousier a customer I am. If I contested regularly, I'd be even a worse customer for them at this point. They actually need a fair amount of turnover in DXing, or at least new blood, to make money. We old farts are somewhere between OK to lousy to money losers for them.
Separately, I've been told by a couple of sources who are well placed that a lot of LOTW's infrastructure is in C++. You can't take that to the bank -- I don't really know. But, based on my sources, I believe it to be true.
C++ remains one of the most venerable and popular languages out there. If you have those skills, you can get a job, believe me. I've made good money writing C++ code over the years. Still, it is also one of the most complex and arcane languages out there. It would not be my first choice for a language that is nimble and capable of rapid deployment of new function. But nobody wants to get rid of working code and pay to re-write it. Re-writes happen, but that is deadweight expense no matter how rich the company. Usually the old code is left around and the new code is added on top. I worked with FORTRAN (believe it or not) around 2010 because nobody in my company wanted to kill this core of very old, but critical code. However, I don't think ARRL is doing anything other than in C++ today. Not sure why (or even if it is true). If they do stick with C++ (and they could see it as a net advantage somehow), I would predict slower deployment of new function than we would otherwise see.