> The other one was that we started out on hardware servers. This was right at the early stage of cloud VPS services catching on and becoming mainstream, and we were just behind the curve. It was an expensive mistake.
I wouldn't think 2013 was early stage of cloud VPS services - but maybe I don't have a good enough memory of the time.
It wasn't, technically, but it was in terms of mainstream availability. Digital Ocean started around the same time as us, for reference. So it cloud VPS was just starting to become the popular option, but it was still far from being the obvious default option.
The other reason we went with hardware that I didn't mention was that we wanted to cram a whole lot (thousands) of node apps onto every server and we figured being on a bare metal setup would be an advantage to optimising that. (In practice: Not really, as it turns out!)
I wouldn't think 2013 was early stage of cloud VPS services - but maybe I don't have a good enough memory of the time.