You’re talking about a single record which will be orders of magnitide lower than the TTFB. There’s really no reason this should influence who hosts DNS with modern infrastructure.
I’m sorry, what? At first I wondered if you meant it doesn’t keep anyone else from using DO nameservers for their domain, but re-reading it’s not what you said…
That suggests a fundamental lack of understanding of TTL - it might be easier to think about it as caching. After an update, clients which haven’t requested it since the last TTL window server will resolve it immediately. In the case of DO and their 3600 second TTL, a DNS server which requested the record 60 seconds prior would be 59 minutes behind on the change. To users that’s ~an hour of downtime.
Compare that to the default TTL at Cloudflare, where the maximum blip due to DNS propagation is going to be five whole minutes.
It’s not a question of changing DNS frequently, it’s about adding a huge unnecessary delay that impacts service. You’ll encounter it in any migration whether they’re planned or otherwise.
On the note of Cloudflare I did compare their DNS performance to DO briefly and consult DNSPerf, they’re a good third quicker in most cases- but as I said it’s a meaningless metric.