CFR was something like 1% early on, and higher in Wuhan where there was a partial collapse of health systems. World population is >7 billion, so 1% gets you 70 millions deaths.
If hospitals collapsed and treatments hadn't been found, a not inconceivable 2-3% fatality rate would get us to "hundreds of millions".
(This estimate is unrealistic in other ways though, e.g. communities would begin imposing their own and hoc lockdowns if hospitals were overwhelmed, and global spread might have been slowed.)
Non-random samples get you to up to 10% fatality rate. I have the data for my region and in march-april 2020 the IFR was 10-15%. Of course, that was due because there weren't enough tests and only the ones that were really sick entered the statistics. But for the original strain, IFR was around 1% and that number was pretty consistent among a broad number of studies.
Also, take into account the amount of diseases that are getting late treatments due to healthcare saturation. I don't know of any formal studies but you only need to talk with any doctor to see that they're consistently seeing diseases in later stages and therefore harder to cure.
If hospitals collapsed and treatments hadn't been found, a not inconceivable 2-3% fatality rate would get us to "hundreds of millions".
(This estimate is unrealistic in other ways though, e.g. communities would begin imposing their own and hoc lockdowns if hospitals were overwhelmed, and global spread might have been slowed.)