I think the modern consensus is that reasonable rate-limiting is a superior option. It doesn't allow for DOS, but still prevents brute-forcing attack. You can even set things up such that too many attempts triggers a notification that something is probably wrong, but users can't be locked out of their own accounts.
"Allow 5 tries, then lock everyone out for 10 minutes" is a rate limiter. If you're talking about an arbitrarily complex one: either you're able to be locked out too, or a botnet allows near-arbitrary retries. A botnet of insecure cameras, perhaps.
It's all a tradeoff. But a smart rate limiter is complex, and rarely necessary when you can be expected to have physical access to the device. Plus, these things haven't even managed basic security, why should we expect them to implement a good rate limiter?
