Pin 25 is a supply rail for termination power. Both Macs should have identical voltages to prevent power supply problems.
I don't *believe* old Macintoshes provide termination power, and I seem to recall it being necessary to ensure that one and only one drive/device on a SCSI chain provided it. It's been long enough I may be mistaken, however.
Now, I'm a little rusty when it comes to parallel SCSI, but... so far as I recall, and as much as it pains me to say it, I sort of doubt you'll *damage* anything (by that I mean "electrically") if were to try this, at least with Mac Pluses. The arrows on that pinout diagram are somewhat misleading, as SCSI is essentially a peer-to-peer network. The thing is that as implemented by Apple the computer is *always* the initiator and the peripheral is *always* the target. (IE, so far as I know no on-motherboard 68k Mac SCSI controller actually allows device "bus mastering".) Which *basically* means the drive should never send a command which would "wake up" both Macs at once*. All transfers, whether read or write, will be "initiated" by one of the Macs, and probably the worst thing that could happen would be if both of them tried to do it at once. However, even then, the fact that the "global" /BSY signal should be asserted as long as the previous transaction is occupying the bus *should* prevent the other Mac from starting its transaction until the bus is clear. Thus... it is my "fairly certain" belief that it's unlikely you'll actually seriously "short anything out" by doing it.
(* The one exception I can think of is if you have a "slow" SCSI target like a scanner or tape drive attached, and the software driver supports using the "Reselection" protocol phase. A device sending a reselect ping to the host *might* wake up both Macs at once. I don't know if the Mac SCSI driver and hardware supports reselection... I'd be sort of surprised if the Mac Plus does, honestly, given the way it basically bit-bangs the SCSI controller. But again, I don't know.)
But yes, if you run both off the same file system you *will* make hash out of it. The fact that the bus arbitration circuitry will *probably* prevent you from shorting out your SCSI controller's data pins will do nothing to prevent both Macs from making mutually-incompatible changes to vital filesystem structures in sequential transfer events.