As with https, moving forward has a cost, but at some point, the cost of staying compatible with old protocols is higher. Before the switch, you have to deal with training staff for the new tech. Once the switch starts happening, you need to train the staff to deal with the risks of keeping the old tech.
For example, I find it easier to have an https-only web infrastructure. Too bad for old deprecated browsers who cannot support proper encryption (and therefore become a security risk for me, and a waste of time to support). Dual protocols (v4 and v6) take more time and money to manage. Initial training can be harsh. Some first-gen hardware was terrible. However, now that it's done, I can't wait to switch off IPv4.
I thought that was like the Voyager leaving the solar system story that keeps repeating every few years but never seems to actually happen.
Only meant partly in jest. When they start to get really rare, the price of IPv4 blocks will keep going up, until some institutional holders of large blocks start selling. It could take quite a long time before it's really gone for good.
Because while it might be a lot of effort for the large institutional holders when they've historically allocated stuff assuming they have the whole huge block, eventually the price-cost trade-off will reach a point where the address space is valuable enough.
ISPs are already dropping support for "real" IPv4 and only offering NAT. At the other end, there are already cloud providers that charge extra for an IPv4 address on your server.
A couple of years ago my ISP dropped their usenet servers - not enough people were using them to justify the cost/maintenance effort. I expect we'll eventually see the same thing with IPv4; not for ten or twenty years perhaps, but once "everything" is using IPv6, it becomes a cost center.
Look further down in this thread and there are already users in Germany being signed up on DS-Lite. Their only IPv4 access is behind carrier-grade NAT. Anything that wants to make a direct inbound connection and doesn't want to proxy (VOIP, games etc) will need to support IPv6
I imagine people will start deploying multiple HTTPS sites on a single IPv4 address but with individual IPv6 addresses. People still using only IPv4 on Windows XP with Internet Explorer will not be able to access those sites (no SNI), but they will soon be a tiny minority.
It feels like it'd need to be a dramatic cost saving in order to kick in at 50% adoption.