Quick explanation, they hold the data sent from the cpu before it gets to the 2A03. There are a few other components that control signals to those chips as well, like 2D which should be a LS138 appears to clock the data in and out of 2H and 2J.
Well this bit of information came in handy two years after the fact. I was playing SPO last night when the sound (which has been working perfectly since I fixed it two years ago) suddenly went screwy in the middle of a game. The announcer was doing the count and then got stuck in a loop doing the count, repeatedly counting from 1 to 10 but in an altered tone of voice. The graphic portion of the game continued fine, but there were no sounds, aside from the announcer counting.
I restarted the machine and the wrong music came up when I powered it on, along with constant crowd noise which wasn't supposed to be there. I coined it up and the sound for that was gone. The music when you press the button to start was gone. Most of the gameplay sounds were gone or incorrect.
I tried various things like swapping the sound CPU's with known good ones, as well as the ROM that I was told holds the sound data, and that didn't help. I was about to remove all the SPO stuff and use it to convert a different, working PO board when I remembered this post. At the time, because of 2600's post here, in addition to the pair of LS374's I ordered for 2H and 2J, I also ordered a pair of LS138's for 2C and 2D. I didn't need to use them so I stuck them in the cupboard.
So I figured I'd give that a try. I replaced the LS138's at 2C and 2D with the new ones I ordered 2 years ago (along with sockets), and that did the trick. My SPO board is working perfectly again, and was only down for a couple of hours.
It was also the first real repair I've gotten to do with my new [to me] Metcal, which was nice.