I'm reading Maxim's datasheet, seeyou're looking at a sheet from another manufacturer that may change, leading to the same kind of incompatibility as seen with the MAX232 variants.
Not recently but over time, yes. From memory the initial spec was+/-12V. I think it was RS232C that lowered it to +/-9V but don't hold me to that. It was either RS232E or F that lowered it again to +/-5V. Even the 'F' revision is fairly old now, but 'C' got so deeply ingrained it's not unusual to see references to RS232C even today. All specify inputs must tolerate up to +/-25V so there is cross-compatibility in terms of avoiding damage at least, but RS232 is one of those standards that often doesn't get implemented rigidly, e.g. power thieves, discrete implementations that are not strictly compliant, and so on.
Acknowledged. MAX232 is the classic part for this application and where I suspect most people will have gained their initial experience. It's the one that has the most issue with the cap sizing. It's also a 5V only part. The MAX2323 is a 3.3V/5V part, at 5V it's interchangable with the MAX232 but the latter has higher output levels. If thresholds are the marginal factor here that's why I suggest it as a possible experiment.