I've been testing Pulp's integers and they appear to be signed 32 bits. This means the integer range is -2147483648 (minimum) to 2147483647 (maximum).
So if you're trying to encode data in an integer to have no risk of overflow you should use no more than 10 digits for binary or 9 digits for base-10.
Wanted to share since I didn't see it documented anywhere 
Interestingly, the known modulo strategy breaks far below the integer upper bound 
Given y = 10, it seems to work on x = 41943058 and break on x = 41943059. Tested with Playdate Simulator.
This seems like a bug with /= or floor. I'm not confident on which because for both 41943058 and 41943059 the log output of /= appears as 4194306.0 (which seems incorrect) but the floor is correct on 41943058 only. I suppose floor seems more likely.
As far as I'm aware, encoding data in an integer is not feasible without modulo as it's serves as the read mechanism. Factoring this in reduces the numbers in my previous statement to 8 and 7 digits.
PDX zip file: Modulo-Testing_20230204215727.zip (32.5 KB)
Logs:
mod: 41943058
x: 41943058
y: 10
mod /= y
mod: 4194306.0
floor_mod = floor mod
floor_mod: 4194305
mod -= floor_mod
mod: 0.5
mod *= y
mod: 5.0
round_mod = round mod
round_mod: 5
mod = round_mod
mod: 5
---
mod: 41943059
x: 41943059
y: 10
mod /= y
mod: 4194306.0
floor_mod = floor mod
floor_mod: 4194306
mod -= floor_mod
mod: 0.0
mod *= y
mod: 0.0
round_mod = round mod
round_mod: 0
mod = round_mod
mod: 0