Hello,
I wrote my own brainfuck transpiler with ti's own optimization passes and I ended up doing some of the similar optimizations you do

https://github.com/ExtReMLapin/fast_brainfuck.lua/blob/main/README.md

For the following optimization I use an Intermediate Representation system as it's much easier to work thru the multiple optimization passes.

Assignation pass:

You already do the clearing optimization.
If then look at the following brainfuck instruction it can become an assignation optimization as the next one could be for increasing the value of the current cell.

That's one step ahead of the clearing optimization pass. It doesn't give much performance boost but you'll need it for what comes next.

Memset pass.

As you might end up with multiple assignations with the same integer (not increases but assignations , see above!) you can have one cell shift/tape position move between each assignation.

Which means you'll set the next N cells to the I value.
I turn this into a memset optimisation call :

Example using lua pseudocode :

Also i'm not sure we're doing exactly the same thing on the loop multiplication unrolling, feel free to check my code on my repo