Unless you absolutely can't use some simple twisted pair cable for this, I see no reason not to use RS-485 transceivers anyway. Problem solved for a couple bucks.
If you can use twisted pair cable, then you can also use singled ended signalling. There *are* 10 Mbit/sec and probably higher standards which use single ended signaling like 10-base2.
So what would a practical implementation look like? Assuming a 100 ohm twisted pair (you are going to need a transmission line for this but it will also work for normal RS-232), I would series or shunt terminate the transmitter side and drive it to +/-5 volts on the receiver side. (1) The receiver side then has no termination but also sees no reflection, or rather all it sees is the reflection but that reflection is the full +/-5 volts. Now the cable capacitance which normally limits RS-232 data rate is irrelevant.
But I know of no suitable integrated drives which do that. I have build discrete drivers before which could.
(1) This means that the series termination requires a +/- 5 volt driver with a peak current of +/- 100 milliamps (actually +/- 50 milliamps) and the shunt termination requires a +/-50 milliamp driver with a voltage compliance of +/- 5 volts. Now you know why RS-485/RS-422 used such low voltages.