Compressing working memory is quite expensive in CPU cycles compared to just using it. Identical page merging is also reasonably expensive and only works if pages are, well, identical. Your Electron runtimes are probably not going to be bit for bit identical, not even small sections, due to offsets not divisible by page size. Tuning for better deduplication is possible, but then you're at an awkward mix of caring and not caring about performance.

Compression is pretty fast on modern CPUs so I would disagree about that. But yes depending on the compiler/switches/versions the Electron run times are not going to be bit for bit identical but it's a stretch to say there will be nothing in common - all the dependent library code for e.g.

Woudldn’t that break ASLR?

ASLR applies to virtual memory / process address space. Merging or compression is of physical pages - the page is uncompressed on the fly when the process accesses it. It works because the OS is in control of the Virtual->Physical mappings and can, on a page fault do the decompression transparently for e.g.