I've only used it for mammalian work, but the Cedex HiRes Analyzer from Roche is pretty sweet. From their docs you can analyze particles from 1-90 μm in size (the cells I work with are about 20 μm), and there are a ton of configuration options depending on your needs. As long as your cells can be stained with Trypan Blue (it's been a long time since I've done bacterial work, pardon my ignorance) to look at viability, then this machine should work quite well. When set to Maximum (i.e., the analysis takes as long as possible), I get results in less than 4 minutes per sample - and this includes about a minute for the sample to settle in the flow cell. If you lower the sensitivity, you can get results faster.
Basically the way it works is you take 300 μl of your liquid culture (diluted if you think the concentration is too high), and add it to a sample cup. Put all your cups in the tray, fill out the sample info and analysis options in the MultiRun screen, and hit Start. The machine includes all the reagents onboard, so it mixes the Trypan Blue with your samples, then injects the proper amount into a flow cell. The sample is allowed to settle, then a high-resolution image is taken of the entire length of the flow cell. Then, depending on your sensitivity settings, a certain number of frames are taken from the full-length image and analyzed. An average count is then generated based on all the images, and your final numbers are spit out. You can then print or save a report of the cell counts and viability.
The software comes with some analysis tools as well, so if you keep multiple cultures going over time with different growth conditions, you can aggregate the data for each one and generate graphs of cell counts, viability, cell size, etc. - there are at least 10 or 20 parameters to choose from. All the image analysis options are configurable, so if you look at the results and think the machine is picking too many dead cells, or not enough, you can change the parameters to your liking, and save the options as a custom analysis type.
One option I like that would beat the pants off CFU-based counting (besides the drastically-reduced time) is the fact that the machine can detect aggregation of the cells. One issue with CFUs is that it's hard sometimes to know that a certain colony really arose from just one parent cell - if you let the colonies grow long enough you may be able to tell based on size, but it's a tough call. Since the Cedex is looking at the cells themselves, it doesn't matter if they're aggregated or not, as they'll all get counted (assuming the aggregation particle doesn't contain a huge amount of cells, like 50+, that can't be resolved by the software due to occlusion of the image).
I don't know how much these guys cost, so depending on your budget you may not be in the market for one, but if you're going to be doing a lot of counting then one of these is invaluable. I'm sure other companies make similar instruments, but I haven't used them, so I don't know how they compare to the HiRes. I do know that the person at my company that was responsible for buying the HiRes is very good about researching possible alternatives, so this one is probably as close to the top of the line as we could find, for the needs that we have.
Good luck with your counting!