I remember chatting with someone intimately familiar with k8s and docker. We were talking about an app I was working on which deployed to heroku. He asked how many dynos and I told him (it was under 10) and he said: "yup, you'll not need k8s".
10 dynos and a big database can serve an awful lot of users.
I've been in meetings where the combinded cost of the time taken to discuss if we should use a thing was more than the cost of the thing. Utterly infuriating.
Even when I worked for an agency there seemed to be an automatic discounting of the cost of people's time vs spending actual cash (and cashflow wasn't an issue).
I appreciate that there is often good reasons to DIY, but when there are not I will always favour something off-the shelf unless it is significantly more expensive.