I have a platform I made for my own use with the idea of potentially opening it out to a wider audience later. It's always been on Linode and will be there initially but I can see myself migrating it to Elastic Beanstalk on AWS if it reaches even a slight demand.
I tried to future-proof users being able to use their own, registered domain instead of a subdomain I provide them (user.example.com). As a proof-of-concept for myself, I had my own user domain (me.com) registered separate from the platform domain (example.com) and was able to make this work perfectly by creating an A Record on the registrar for me.com that pointed to my Linode IP address and created a virtualhost entry that caught all domains pointed its way and verified it was a registered one for a user.
So, all working.
I became aware not to long into the project that this isn't an ideal practise. Redundancy wasn't a major concern as, if my single IP went down, the service was down anyway. Having 5 other name servers to fall back on wouldn't help in this scenario. My major limitation was the liklihood of the IP address changing. I even envisage the change to AWS in the future, as mentioned above.
My question is, is the changeover from pointing user's self-registered domains to a method that allows for me to change IP address really going to be as either as financially hard hitting as it appears?
I won't make a lot of money from this platform as users will not be paying a lot to use it (it offers a fresh, local spin on slightly similar services but those services are heavily bankrolled and priced at a low cost) so I am already cutting it tight with hosting costs. I also don't want to take on further responsibilities, such as being a domain registrar and am not savvy enough to confidently run my own name server.
The typical answer to this would be to use an independent name server and register an A record for each domain. My situation is a little more niche where the platform is more built on representation for a profession. So I may have lots of users with custom domains, little traffic and paying little.
Most DNS services seem to offer limited A Records or small DNS query amounts. I have the potential to see a short spike in queries and an unexpected bill for a single month could be financially disastrous. I believe I have this mostly catered for on the hosting side but would now have to worry about it on the name server side too. I can't oversell resources to the degree that a typical platform could as I would have users in the tens or hundreds rather than the thousands.
Although it's certainly not recommended, unless I want to pay through the teeth, is my best option still to go back to using an A Record for my IP Address and request that users update it when/if I change server? I would love it if a company offered a vanity IP address that I could direct to a second one (my server) and change later but I guess nameservers fill that niche for most uses.