It seems like that changed somewhere around the turn of the century, whereby businesses started to decide that it was better to cut down on customer service, and in some cases, go so far as to ban customers. The first cases of this I recall reading about had to do with Best Buy, and specifically their policy of banning people from their stores who made a lot of returns.
I'm not really sure how it ultimately maths out - i.e., whether it's long-term optimal to drop troublesome customers or merely short-term optimal, and this was primarily taken from the perspective of retail.
As such, I'm sure the math changes a little for subscription services. However, I also recall my prior employer's support activity followed a power law distribution across its clients, so it wouldn't surprise me if a policy to drop particularly noisy clients is a net savings there as well.
may be true in the past, but a business cannot scale up good customer service - it's at best a linear scaling, where each new customer costs the same fixed amount due to needing high touch/people to manage that customer.
With the internet, businesses have found scaling to work better by ensuring your fixed costs stay fixed even if you scaled up customer count - this includes support/call centers etc. Without doing this, the business cannot scale up exponentially.