I know someone who needed to run a large-ish distributed workload. The project had a pretty respectable budget (I think more than $10k, less than $1M). They w

I know someone who needed to run a large-ish distributed workload. The project had a pretty respectable budget (I think more than $10k, less than $1M). They were going to rent DigitalOcean servers, because benchmarks showed they were the cheapest hosting provider around. Then their account got banned due to "patterns associated with cryptocurrency mining." And they had no way to get in contact with a human, appeal the ban and explain their situation. (Interesting side-note: Even if they were doing cryptocurrency mining, AFAICT it isn't against DigitalOcean ToS.) They ended up switching to another provider (Linode I think?) Anyway, what I'm trying to say is, it's an example of a cheap provider who has good paper specs and even good benchmarks. But when you actually try to use the platform at scale, you'll discover there's some catch and you can't actually use it. (I speculate that maybe DigitalOcean madly oversubscribes physical CPU, so anyone who actually uses the resources they pay for trips the automated system and gets their account banned, because DigitalOcean can't both keep their cheap prices and pay for the engineering effort to figure out how to throttle customers who peg their CPU. So they make enough money to stay in business from websites and non-CPU-heavy workloads, but have to use the backdoor "cryptocurrency mining" banning excuse to turn away customers who basically want to buy CPU-hours.)