I'm no expert, but certainly have a lot of experience in this area of communications. AT&T runs their 2G systems on 850MHz. (Actually, the full range of the spectrum there is 824–849 and 869–894, but they call that mess '850MHz' and I'm too lazy to locate the exact AT&T Oakland freq there in the FCC database, but it doesn't matter in this case).
The DTRS system Oakland is using (aka APCO-25) typically comes in two different frequency flavors .. one down in the 700MHz range, and the older and still more popular 850MHz which also really isn't exactly 850MHz, either. Oakland's DTRS system (easy to look up) is operating in the 851~855MHz range. As you can see, that slides right in between those two big blocks for AT&T's cellular service.
It takes a decent radio to reject all of the adjacent frequency signal when you're sitting right under the transmitter, and my trusty (and up-to-date) model Radio Shack scanner can't do it. I have to hit the ATT (attenuation) button to bring down the whole incoming signal level a bit. Good cop gear is a lot more selective and also utilizes better quality AGC (automatic gain control) on the RF than consumer gear, so a DECENT cop radio won't be the least bothered by the adjacent channel interference. That's part of the reason that they cost a good chunk of change.
DTRS is digital. Like VOIP for phones, voice sounds are turned into digital packets and turned back into sound at the other end. Unlike the old analog radio days where you could still pick the words out of the noise if you had a good ear, below a certain signal threshold, the digital packets can't be decoded and things very quickly become completely unintelligible. In dispatch-ese, it's called "going digital", as in "Paul 258, you went digital", meaning the dispatcher heard garbage instead of words.
Anyway, unless AT&T was running an 'out-band' operation there in Oakland, and so far, nobody has accused them of that, the problem lies in the equipment that Oakland purchased, and a really poor testing phase that should have uncovered this problem very early on -- before the actual deployment. It's not like there aren't tons of 2G cell sites there.
