The anecdotes from the 80s & 90s when annual PC sales for any given year were a substantial portion of the overall PCs in use at the time are interesting. The realization that software could be written for the newest machine and it would run everywhere soon enough is not directly applicable to the smartphone market. I think that out of collective habit, many apps assume that building for this year's flagship models is the best course. It may be, but it effectively means your market is much smaller, especially for free, network effect apps.
These apps and their ecosystems rise and fall fast. Tablets & smartphones are spreading downmarket with devices being sold today @ 10% of the first androids.
A $50 phone won't run everything, but it will run some things - useful things.