Should all the relations in that paragraph be maths, not code?

That would probably be an improvement, but the only one that causes me physical pain is the one that won't compile.

Every engine and engine adaptor in [rand.eng] and [rand.adapt] has similar relations as code, but I think [rand.eng.mers] is the only one that's a real problem.

Hm, maybe we need more thorough cleanup. I don't like silently conflating different symbols, and w and w are different things. We should have proper introductions that say "here's a quantity w that does XYZ, and the template parameter w has the value w etc.

Once everything is properly defined and bound, we can consider mixing relational C++ expressions and mathematical inequalities, though I'd still prefer if we could just use one (i.e. just mathematical inequalities).

[rand.eng.lcond] p3 would be simpler stated as maths:

If the template parameter m is not 0, the following relations shall hold: a < m and c < m.

We don't need to special case m==0 in both p2 and p3, we can just rely on p2 to rewrite p3 as:

The following relations shall hold: a < m and c < m.

(Yes, we should introduce a and c instead of a and c, but note that p5 and p6 already use a and c without introduction anyway!)

I think we should put blanket wording in [rand.eng] saying that a placeholder p has the value of the template argument that is substituted for the template parameter of the same name, p. That way we don't need each engine or adaptor to say "the multiplier a has the value of the template argument that is substituted for the template parameter a" for every parameter.

Yes, blanket wording would help a lot.