"This is a disk drive array and surely, long-term, disk drive arrays are the new tap libraries and will not store fast-access data or, eventually, near-line data either. Enterprise on-premise arrays will go all-flash; that is what analysts such as the people at Wikibon are saying."
Yeah, about that . . . wake me up when AFAs have a price per GB that comes anywhere near NL-SAS or SATA. Right now, the sweet spot in price/performance seems to be with hybrid arrays which offer bulk storage on fat, slow drives and a fast flash tier with concomitant storage logic designed to maximize the performance benefits of flash. Contrast this type of design with a legacy storage array which has hybrid storage bolted on (e.g. NetApp FAS or EMC VMAX); those arrays do not make efficient use of flash storage, especially for writes, no matter what the vendors say.
"A second worry is that hyper-converged systems are attacking the idea of a networked storage array being necessary at all."
The hyper-converged space is still, you know, converging, with significant concerns about the strength of the overall architecture. Enterprise customers also often have an existing storage infrastructure investment (HBAs, cabling, switches, etc.) and a pre-existing architecture to support it, so slapping in a new, faster array is less disruptive and often cheaper than trying to migrate to hyper-converged. New companies may want to make use of hyper-converged appliances, but I suspect there's a significant addressable market of well-capitalized, profitable companies who would be happy to have a box which plugs into their existing infrastructure and makes everything go faster.
"A third worry is that on-premises data is going to the cloud."
This concern is valid, but I think it mainly applies to newer companies, as above, or older companies which don't have data that matters to them. And who knows, maybe Infinidat is targeting cloud providers as well.
There are multiple large companies still selling large frame arrays and making a tidy bundle doing it, indicating that there's still a profit to be made in this space. Most of the offerings in the space that I've seen have some woeful deficiencies in terms of price efficiency, manageability, and performance, so a fast-moving competitor certainly has an opportunity to steal somebody's lunch.