Have there been changes in Lazarus which expose the properties of objects in frames to the objects since then? By that I mean at runtime.
If it didn't work would I really be telling you that it was the only safe way to do it?
After sleeping on this: there were very early code generation problems relating to frames, but those were fixed in FPC 2.7.1. I was using them heavily- in the context of notebook-style containers- by roughly 2010 according to comments in the project's source, and other than the problems caused by design-time instantiation have had no problems since.
I'd say that the overwhelming cause of "I can't access a frame's properties at runtime" are screwups caused by design-time instantiation.
There /is/, obviously, a possibility that something related to RTTI has (or at least had) problems. The only way to work that out is going to be by writing the mandatory standalone test project, which can be provided in compilable form for others if necessary.
MarkMLl