Posted a reply on github and adding a copy here in case you may not have gotten notification:
Thanks for opening this.
There are several different changes combined together in this PR and I’d like to know which specific ones were necessary to resolve the problem and which ones are precautionary but not essential.
In addition, while there is some free WRAM I am reserving that for other uses, so using a substantial amount of that for the ~2.3K tile cache isn’t the right approach right now. I’ll think about how else I’d like to approach that and may do a different implementation to achieve that.
























