There are a few designs I've been meaning to try to deal with the problem of contamination. Haven't got around to it yet though. Is there a way to cheat a contaminant onto a square? That would make research simpler.
Design 1: If you build top-down, a continually running 2 pump stack will move fluids through a tile-- without that tile ever having fluid at the end of the tick, or ever having fluid when dwarves decide to move. Using magma, this can create a tile that's bathed in magma (potentially every tick) but is not impassable.
I've mentioned this before, and people are still coming up with complicated plans for dwarven liposuction, but this is a simple and perfectly reliable method. A dwarf that spends enough time in a tile bathed in zero-tick magma will lose fat. With an unpressurized reservoir, 0-tick magma is on a cycle of about 30 ticks, and a dwarf accepting a long enough job in that tile (deconstructing a wall, for instance) will eventually lose all of his fat, without job interruption. The cycle can be reduced to a single tick by using a pressurized magma reservoir.
I'm not sure how this would affect contaminants in the 0-tick magma tile. Testing with water would be worthwhile as well.
Design 2: If you build a bottom-up, continually running 2 pump stack, fluid will exist in a tile for a single tick, about every 30 ticks. Normally, this isn't very useful, because fluids can flow in even a single tick. They cannot, however, fall, not on the tick they're pumped. Building a grate (with open space underneath) under your 1-tick fluid square allows you to bathe a tile in a constant cycle of fluid, without fluid loss. It's not totally clear if this will remove contaminants, or if it will interrupt dwarves that take too long to complete a job in the affected tile.