Your question is actually about how fast you are going to spend the fuel of your first stage: Are you using four or five engines simultaneously to convert it into speed? When you use all 5 engines but throttle down to 80% thrust, it's exactly the same as if you would use 4 engines at 100% thrust. To answer this question, we first need to talk about what forces affect a rocket during lift-off.
There are two forces which prevent your rocket from getting into orbit.
- Gravity
- Atmospheric Drag
The first, gravity, is accelerating your rocket constantly downward until you have enough horizontal velocity to cancel it (achieved a stable orbit). The more time you spend affected by gravity, the more speed ("delta-v") does the gravity give you which you have to cancel by expending additional fuel. For that reason it is economically to accelerate early, so you get to your orbit faster and consume less acceleration from the gravity.
But there is also the second force: atmospheric drag. Atmospheric drag depends on atmospheric pressure and speed. The relationship with speed is quadratic. The faster you go, the more speed ("delta-V") you lose through air friction. That means it might not be so good after all to go too fast while you are still in the lower atmosphere.
So where do these two factors cancel out?
The ideal speed to balance atmospheric drag and gravity drag (assuming perfectly vertical ascent) is equal to the terminal velocity on the current atmospheric density. That speed depends on how aerodynamic your vehicle is.
When you go faster than this, you are wasting fuel on atmospheric drag. When you go slower than this, you are spending unnecessary fuel to fight gravity.
To get back to your initial question "what's better: serial staging or parallel crossfeed staging": It depends on your total thrust-to-weight ratio in the lower atmosphere. But my general experience is that a rocket gets higher with cross-feeding.
But what when you are already in orbit?
The truth is, it doesn't matter. The amount of delta-v you get per liter of fuel depends on the total weight of the ship and the average fuel-efficiency (Isp) of the engines you use. It doesn't matter how fast (through how many engines) you spend it. All that matters is to avoid having more mass than necessary (get rid of fuel tanks when they are empty). An orbital transfer stage with less engine power is more efficient, only because it tends to have a lower overall mass. This, however, is bought with longer burn-times to get the same amount of delta-v. Longer burn-times can sometimes mean less efficient burns because you can't hit your maneuver nodes that exactly, but this only matters in situations where a burn is very time-critical.
Unfortunately not. At best, you receive 100% from surviving resources landing on the launchpad at the Kerbal Space Centre. The percentage will decrease rapidly as you get further away you get, down to a minimum of 10% at the exact opposite of the planet.
As for strapping parachutes to SRB's:
To reduce the amount of calculations the game has to preform, all objects over 2.5km from the active vessel are placed "On Rails"(physics is no longer calculated, object continues on last predicted trajectory). All "On Rails" objects that are too deep into an atmosphere are deleted immediately.
It's also worth noting that only objects attached to a command pod or a probe body will be recovered, the cheapest probe body in the game is three times more expensive than an SRB and you don't get much back from an SRB, because most of it's cost is in fuel.
There are techniques that would allow you to over come these problems, but the extra effort and costs required to achieve them are more than you recover from it.
Best Answer
I did what you did, but put a stack decoupler between the Mk1 Crew Cabin and the Mk1 Command Pod, and gave the crew cabin its own heat shield and parachute. The heat shield is enough to keep it properly oriented during reentry without any input. The one trick here is that you want to delay separation as long as possible otherwise the cabin and pod will get too far apart and the one you're not controlling will be destroyed when it goes past the simulation range.
I experimented with trying to make aerodynamic reentry vehicles that could glide to a landing but didn't have any luck. This solution worked well, and I was even able to expand it to two crew cabins separated by a second decoupler.