I've been thinking about that. This rail serves the RS422 ports, driving the negative half of the TX and RX differential circuits. The positive sides of each signal are correctly filtered. So I'm guessing that the filter isn't essential but is there as standard design practice to protect the signal when it is close to limit of specs - say with a very long cable to a printer or modem. It's fairly low current and short duty cycle and the overvoltage on the reversed cap might even be within, or not far outside, the tolerance of the original electrolytics. The ESR generates heat, but the aluminium cans might dissipate this better than the replacement poly tantalums, and so last indefinitely with moderate or low usage.
It could be interesting to test a machine in the original state, just by finger touch on the cap, to see whether the reversed cap is noticeably warmer than its neighbours when in use, say with a printer attached to the port. I don't have one of the relevant machines to test myself.
I'm an amateur at this, and would be happy to hear opinion of a proper EE. :-/
Rick