TITLE: Almost free circle actions on spheres
QUESTION [9 upvotes]: $\DeclareMathOperator{\Fix}{\operatorname{Fix}}$I am looking for any reference regarding the following problem:
Problem: Consider a smooth almost-free action of $S^1$ on a smooth sphere $S^n$. Then for all finite subgroups $\mathbb{Z}_m\subseteq S^1$, the fixed point set $\Fix(\mathbb{Z}_m)\subseteq S^n$ is either empty or a rational homology sphere.
This is well-known to be true when the action is linear, or by Smith theory when $m=p^k$ is the power of a prime $p$.
A stronger statement would be that for every action of $\mathbb{Z}_m$ on a sphere $S^n$, the fixed point set $\Fix(\mathbb{Z}_m)$ is a rational homology sphere. What I know is that when $m$ is not the power of a prime, then $\Fix(\mathbb{Z}_m)$ is not necessarily a smooth sphere, but I have not found any indication about its possible (rational) homotopy types.
What boggles me the most is that, I had imagined this problem to either well known to be true, or well known to be false, or extremely interesting – however, I have not found any evidence for any of these.

REPLY [2 votes]: Back in the early 1950's Ed Floyd and Pierre Conner found that composite order cyclic groups could act rather differently that cyclic p-groups.   In their tradition, the publication [J. M. Kister, Examples of periodic maps on Euclidean spaces without fixed points. Bull. Amer. Math. Soc. 67 (1961), 471–474] shows that the cyclic group of order 6 can act on a big enough Euclidean space with no fixed points.  One point compactifying, one gets an action on a sphere with one fixed point.
There are quite a few survey papers on this subject. Here is one that I found quickly, that has lots of references: Reinhard Schultz, Bull. A.M.S., volume 11, October 1984.  See also lots of other papers by him, and also Bob Oliver.