The important point is that propellant is cheap, cheap, cheap, and loss of a launcher is as expensive as all hell.

Best reliability calls for a completely reuseable single stage launcher, with engines of such a size and number that, if at any
time one of them must be shut down, the others throttle up to compensate, and you just keep going. Use of two parallel stages is
a distant second because separating two vehicles at high speed in the atmosphere is not simple. (A two parallel stage launcher
can be designed in which separation occurs outside the atmosphere; this does make the return of the first stage a bit more
difficult.) Series staging is terrible because you take off without the second stage engines running, and thus without knowing
whether they will start and run correctly. In any case there's little reason to use multiple different engine designs in one
vehicle.

The "reusable" part means that you can get all of the design and manufacturing errors out of every flight article before it goes
into service. This also means that, in service, the chances of a catastrophic engine failure are negligible.