For a start I would just try not using cache mode full, because this is a mode that is generally not recommended for most uses. It will force all files to be downloaded completely before access - thus disabling any capacity to stream or quickly access things. It may be appropriate to use in very spesific setups, but you should probably prefer to use --vfs-cache-mode writes
Secondly, while it should work in theory I suspect that using a single-file encrypted container like that will be quite slow in many regards. rclone will need to do a lot of seeking and jumping around in that file - and each such operation takes some time to do. Using rclones native crypt support is almost certainly going to be much faster - so I would at least consider doing that. Let us know if you need help regarding that or want to know more about it (and you can read up on the documentation as well of course).
There are some limitations on what kind of file-access is possible to do in practice through a remote. write caching should normally fill in the gaps though.
So try cache mode writes and see if that works first of all, then report back and we'll continue from there.
I have no experience with luks, but that error indicates it is measuring the size of the file and finding a problem. I guess it is mis-measuring for some reason and maybe is effectively polling size of the rclone file-segment or something instead pf the actual because of the method it uses to ask. That is my best hypothesis based on the little information we have to go on here.
Maybe @ncw can give you a suggestion for what settings it might be worth experimenting with.
As a shot in the dark, I'd try maybe setting
-vfs-read-chunk-size 1G
or a size that is at least the size of the file...
I don't know if you can completely disable read-chunking (I have never tried) but setting it to 0 would probably do that.