"Can I really upgrade one of the C9500 without disrupting operations of the other?"
I've no experience with current VSS technology, but in the past upgrades usually were more "troublesome" vs. a chassis with dual sups.
Of course the issue if you only have a single chassis vs. a pair, the single chassis, itself, is a single point of failure (although chassis failures are pretty low probability).
Often overlooked with VSS, the IOS, itself, might also be considered another single point for failure, but unlike hardware, its MTBF and MTTR are more difficult to estimate.
In my last company, for cores that only had L3 connections, we ran them as independent platforms using ECMP. Doing this, such a pair could run different IOS versions and/or be different hardware platforms.
For distribution or access we often then would use redundant sup chassis and/or stackable switches.
The above, for non DC. For DC, we would use Nexus and its vPC, and FEX, which can be also a challenge to update, although I recall (?) didn't run into issues, as often, keeping downstream devices on-line.
