In RS232 a valid "spacing" voltage (data bit 0) is +3 to +15 when connected and tolerant to +25 when open-circuit. The "marking" (data bit 1) voltage levels are -3 to -15 and -25. The drivers and receivers are tolerant of a short to another signal. The signal will not go through then, but the drivers & receivers should recover once the short is rectified. Suggested reference
Since -3 to +3 is signal's non-man land, many receivers use a biased mid-point away from 0.0V to 1) prevent a grounded input from chattering 2) or bias to a mid-point like +2.5V to allow 0 and 5 voltage signals.
RS232 also defines a maximum slew-rate (dV/dt) that is generally ignored else high baud values would not be allowed.
Agree with @Wouter van Ooijen. It worked then and many technologies are evolutionary - not revolutionary.
Some additional musings.
Some RS232 terms go way back to early telegraph.
The terms for signal condition of "Space" (Data bit 0) and "Mark" (data bit 1) denote when a paper tape was "marked" or not. Soon this was changed to audio signals instead.
RS232 "break" condition (All data bits 0 and stop bit 0) corresponds to a telegraph operator "breaking" the circuit. Back-in-the-day, all the operators, shared a signal wire. When not sending, they left the keypad in a closed position - creating an idle voltage. Any one operator signaled "break" to announce a following message. By then tapping on the keypad, the circuit momentarily created the dits and dahs of Morse code.
Moving forward with teletype, I suspect a dual +/- voltage was needed to drive the mechanisms closed and open. But the concept of a "break" to announce an incoming message persisted. (This does not address why it is +V means 0 and -V means 1 versus the other way around.)