Generally they do not do a micro system-by-system comparison and then try to integrate them into a best-in-class Frankenstein sum of the parts.
On first pass it may seem short sighted and that they may lose good tech, but given the 1000s of systems this exercise would take forever.
Further due to all the dependencies, you really can't pick and choose anyway.
In this particular case it would also be pretty hard to argue CS is better that anything (other than losing money) than UBS anyway. CS has been on deathwatch for the last 10 years, at least..
I’d imagine it is probably like merging two spider webs together.
If you come to one of the big banks and just look at the technical components, you can delineate the history of each merger/acquisition from that, going back 30-50 years.
A thing that shocked me: a lot of systems have a weird nationalistic twist to them (at least the old ones), so if you work in non-US, you will see systems that do not speak english in their API, but french, dutch, german, etc. all in the same bank. This is...problematic, especially since they usually use weird text encoding and not something like UTF-8.