I didn't know your usage level! For small projects they've served me very well.
For larger projects then I'd rather host my own anyway, as I want pretty complex rules for handling queries from different regions, failover, number of IPs to return, and so on. We do exactly that with gdnsd and handle about 450m DNS queries per month.
The reason Route53 excited me is it has failover, healthchecks, etc. built in in a way I can sleep through things failing without anything breaking.
I don't actually use 2 million DNS queries / month, yet. But if I implemented DNS-based failover and shortened the TTL from the 12 hours I have now to something like 5 minutes? Ya, that'd clear 2 million pretty quickly. ;)
For larger projects then I'd rather host my own anyway, as I want pretty complex rules for handling queries from different regions, failover, number of IPs to return, and so on. We do exactly that with gdnsd and handle about 450m DNS queries per month.