Shouldn't be an issue. It sounds like you have a very small network, with that I would consider your 'choke' points. I have run discovery on 6-12 /24's without issue, but most of the network had 10G up links, some access layer 1G and all dual homed. If you have any single homed devices, or chained connections you might see a slight issue/delay. A Good amount Cisco Gear is designed to drop SNMP before it adversely affects the running order of your switch/router; so if you network hardware is 'up to speed' then the delay I would expect you to experience is longer page load times(SW) while the discovery is running.
And in regards to sizing, anything more than a /24 usually takes a long time to get through. I like to keep things /24 or smaller if discovering all addresses in a block.
How much bandwidth you use is really dependent on how many devices there are in a given subnet and what protocols they answer to, but it will always be pretty small. To get a sense of scale, I checked in on a server I used for a discovery earlier today, we tested 900 addresses while still doing our normal monitoring of several hundred nodes and comparing the peak of the discovery against the "normal" baseline network utilization looks like it added 3-5 mbps of bandwidth utilization for us. Your results may vary but that is the ball park you are dealing with. If you are trying to scan a /22 on the other side of a t1 it could become an issue, but for anything remotely modern it would be a blip on the radar, easily less than the average youtube utilization on most networks.