I had no idea just how level epoxy levelling left the surface so I checked one of my rails this evening. About 1750mm long, joined to the opposite rail about 1000mm away via a single central bridge. Epoxy around 4-5mm deep (there was a dip in the middle of the 100x50 box section rail support of about 1.5mm). Once set, I removed the meniscus from the epoxy with a coarse file and also ran the file over the top of a few bubbles that had formed.

For checking, I used a precision level plus feeler gauges, resting on a couple of ball carriages on the profile rail bolted directly on top of the epoxy. Using a 200mm level, I measured the deviation from level at 200mm intervals along the rail, then calculated the variation from a straight line between the two ends of the rail (which were about 4 thou different in height. This was because I had moved the frame slightly from when the epoxy was first poured.) I'm sure that there was plenty of scope for errors to creep in, but at least it gives a feel for the overall magnitude of the errors. I'm sorry for the mixture of imperial and metric units in this post but I still can't get a grasp of the magnitude of small measurements once you are down to tens of microns and I have to go back to "thou".

I was surprised by how good the results were. Remembering that these figures are variations from a straight line between the two rail ends, the maximum variation from "flat" was about 7 thou corresponding to a dip in the middle of the rail and smoothly rising towards each end. Given that this resulted from an epoxy self-levelling along a long, narrow 1750x30x5 "channel", and even though fairly runny by epoxy standards it's still not quite like water, that seems pretty good. Knowing the numbers gives me more confidence in moving forwards, and I'm planning now to use epoxy levelling on the gantry as well.