I think this is my favorite answer:
To determine how much free space a system requires, one must account for two variables:
The minimum space required to prevent unwanted behavior, which itself may have a fluid definition.Note that it’s unhelpful to define required free space by this definition alone, as that’s the equivalent of saying it’s safe to drive 80 mph toward a brick wall until the very point at which you collide with it.
The rate at which storage is consumed, which dictates an additional variable amount of space up be reserved, lest the system degrade before the admin has time to react.
First let’s assume we can disregard point 2, because once the drive is full we won’t be writing any more data. In fact, most of the other answers in that thread can be ignored for the specific use case of farming Chia.
For the first point, let’s define unwanted behavior. Unwanted behavior would be defined as poor farming performance, specifically being unable to answer challenges in time because of slow disk access. If you read through the rest of the answers, poor performance is mentioned when writing new data to an almost full disk. Because there isn’t a large amount of free space, contiguous data might not be possible, so the OS must fragment the data in the smaller empty spots on the drive. Again, this won’t be a problem with Chia plots because they are huge files that are written sequentially to the disk once and therefore will not be fragmented. So: at this point our theory is that Chia farming should not be affected by these performance concerns. Let’s test it!!
Luckily I already have, being the author of that thread you mention. I’ve been running over 60 SATA disks with this method, all almost completely full except for a few gigabytes and all connected through port multipliers and PCI add-in cards. Performance is still excellent. In fact, I’m not only farming Chia but also 5 other forks, all at once. They all live in their own virtual machines and farm the same plots from the host drives. Even with the equivalent of 6 farmers hitting the same plots at once, I have no performance issues whatsoever and consistently win proofs (mostly on forks lol)!