Larger cap's tend to have larger ESR and ESL. That's particularly true of electrolytics, but applies somewhat to tantalums as well. You probably would want a large swamping cap, ~10-33 microfarads together with a 1 nf chip cap, which should have VERY low ESR and ESL. If you're really fussy, as I am, you might throw in a 0.47uF cap, too. What you do on the side remote to the MAX232 is somewhat more discretionary, as you have to consider the effect on the entire circuit. If you have, say, 51 ohms of total resistance, with, say 33 ohms on the Vcc line and 18 on the Gnd, then Gnd will move around quite a bit, but not enough to damage input sensing, and certainly not enough to move output levels to an illegal or erroneous state.

The output level of the MAX232 into 300 ohms is +/- e volts, which is no longer a valid level, so I'd guess that it will drive a higher impedance to a valid +/- 3 volts. The short-circuit current is 10 mA. To source that into a load it has to get it from Vcc and Gnd, and running 10 mA through even 75 ohms would, with two short circuits driven, generate a worst-case net offset of 1.5 volts. A short circuit is, of course, no problem for communication, since, when that occurs, something is broken. The Gnd of the MAX232 will wander around a bit, as will Vcc, but it (the MAX232) is generating +/- 10 volts ... with a requirement for +/- 5. I'd guess that the voltage will be changing, on the average, about 1/2 volt under the worst-case conditions. Now, that's just over the thumbnail guessing ... but you see where it leads.

Those resistors are supposed to localize the effects of the charge pump on the Vcc and Gnd to the Vcc and Gnd of the MAX232. I suppose there are better ways, but no less painful ones.

RE