There is a reason it's popular: it's extremely useful for software development to a) have actual hard data on how your software is being used, and b) have a large selection of crash data for debugging rare issues. If it's not the software engineers who want it, it's the technical management who see the immense value in having it.
You're forgetting the monetary incentives. It allows companies to collect personal data of every user and sell it to "our partners" to build larger marketing profiles.
This depends on what kind of data they're collecting. The most common kinds of telemetry data is not actually particularly useful for that, and usage of it for selling advertising, especially to third parties, would be contradictory to most privacy policies (now how much you trust that they are actually following their own policy is another matter: and dropbox does call out that they may try to use this data to upsell you on their own products).
Nonetheless, the potential is there and GDPR does consider it personal data from the point of view of consent, so dropbox is almost certainly violating the rules here even if they do not sell the data for advertising (as unlike the actual data they store, it is not necessary for providing the service, merely useful to the company for improving their service). Such telemetry almost certainly requires an opt-out, and most likely should be an opt-in as far as GDPR is concerned.
There was an opt-out telemetry proposal in Go [0], which caused a huge backlash. The proposal authors were so focused on the benefits of the telemetry, that they did their best to invent all kinds of very convincing arguments why their telemetry is okay, useful, not intrusive, etc. etc. They completely ignored the ethics of the problem - that they are not entitled to users' data without consent.
It took a very dramatic reaction from the community to convince them that adding opt-out telemetry without users' explicit consent is a bad idea, no matter how "non-intrusive" and "helpful" it is.