I don't think it does everything desired by those issues, but it is a fine idea to have this.
I think the documenation should make it clear whether it's "instructions that were left prior to this call" or "instructions that are left afer this call".

Reading when shared.Cpu.instructionsPerUpdate gets incremented, it appears to get incremented after executing an instruction, so when the current instruction is evaluating the suffix and putting the answer on the stack, it's seeing the value minus one.

I think we can just move the increment line from CPU.cs to just before, rather than just after, the ExecuteInstruction and it won't mess anything up. It would make the result of this easier to nail down.

i.e. currently it does this on line 1505 of CPU.cs:

                    executeNext = ExecuteInstruction(currentContext, doProfiling);
                    ++InstructionsThisUpdate;

and it could be changed to this without introducing a problem, I think:

                    ++InstructionsThisUpdate;
                    executeNext = ExecuteInstruction(currentContext, doProfiling);

This would mean that semantically "InstructionsThisUpdate" is now counting how many instructions have happened including the current one, when before it was counting how many instructions have happened excluding the current one.