The distance limit is 100% timing based, a factor of the propagation speed of the signal down the wire along with the Ethernet spec requirement for how long a device listens for a clear "gap" on the wire before transmitting.
Within the scope of twisted-pair cable, a "better" cable is not going to change the laws of physics about how fast a signal travels down copper (we're talking about all-copper cables/wires here, the standard for any halfway decent twisted pair). In fact "better" cables are sometimes defined by twists per inch, where a "better" cable could have more twists, and therefor slightly lengthen the cable, contributing to the problem.
As you exceed 100M, you increase the possibility that a device at Meter 0 starts to transmit at JUST the same time as a device at Meter 95, thereby causing those two packets to "collide", making both devices need to back-off, wait, and re-transmit. These collisions are expected to a certain degree and are part of the Ethernet spec (CSMA/CD carrier-sense multiple acces / collision detection). You can have a fair number of random collisions and still get your "wire speed" for total expected throughput. But if the number of collisions increases too much, you're spending all your time trying to retransmit the same data, and your effect throughput drops considerably.
Switches reduce your collision domain (basically a segment of network where two devices have the potential to collide with each other) greatly because they create a series of sort of mini-networks, between the switch and each device connected to it. Your device now is mostly only contending with the switch for network acces, and not contending with EVERY other device on the network simultaneously. It's more complex than that, but that's the basic concept. Switches started becoming more affordable and more common about the same time that "better" components and cable started becoming more common, so I think a lot of people attribute functioning 120 meter cable lengths to installation more than switches.
MOST switches and devices WILL work at 120 meters for example, provided the installation wasn't done by a half-blind chimpanzee who tried to invent TIA 568-D on the fly (haha, cabling humor there). However, SOME switches will look at the timing of the link and refuse to establish a link if the propagation times are too long. I've mostly seen this happen with enterprise-class switches, but haven't dealt with over-length segments enough to really fully test this. So, the 100M limit is mostly not a "brick wall", until that time on a new job where it is and you're going crazy trying to figure out why your cameras and cables that worked to 120 meters before suddenly aren't doing so now.