It's funny that you bring up the example from The Empire Strikes Back where the Star Destroyers collide, because they are clearly not able to evade even though they have quite a bit of time to do so. We can also assume that the reason the Star Destroyers take no hull damage is that they were shielded.
If you're restricted to differential thrust for turning and do not have the ability to reverse the thrust on one or more thrusters, you have to go forwards to turn. The Star Destroyers that appear to have collided in
The Empire Strikes Back already had a significant velocity towards one another and were rotating towards one another as part of a maneuver to pursue the
Millenium Falcon at the time when their crews became aware of the possibility of a collision.
Spoiler
This is a significantly more difficult scenario in which to avoid a collision than the scenario in Rogue One, where neither Star Destroyer has any significant velocity relative to the other and the vessels are oriented in such a way as to make moving forwards while turning something which tends to move the vessel out of the region where a collision is likely.
In terms of the Executor scene in Return of the Jedi, you didn't take into account the gravitational forces of Endor, Endor's moon, or the Death Star itself, and since we can't see the actual alignment of most of those bodies, it's almost impossible to figure out what the capabilities of the Executor actually are.
The distances involved are great enough that the gravitational attraction that the
Executor feels towards any of the natural bodies in the system are unlikely in the extreme to amount to a net effect of more than perhaps a gravity or two of acceleration towards the Death Star, which is hardly worth accounting for when the constant-acceleration model's estimate puts the acceleration experienced at 150 gravities. The Death Star itself is the only large body both close enough and potentially massive enough to have a gravitational effect worth worrying about, but running numbers for the Death Star runs into a lot of issues - the size of the Death Star is unclear (the current canonical diameter of 160km doesn't really work for what is depicted in the movie, with the most egregious example being the scene where the
Executor crashes into the Death Star; the geometry of that scene suggests that the smallest possible diameter for the Death Star is about 1000km, if the
Executor is taken to be 19km long), the mass of the Death Star is completely unknown, and the distance between the Death Star and the
Executor just prior to the
Executor's crash dive is unclear but in excess of the
Executor's overall length. We can maybe put an upper bound on the mass of the Death Star if we make some estimates of the Death Star's orbital altitude because we can see that the Death Star does not appear to be directly over the point on the moon's surface where the Rebel infiltration team is and does not appear to be massive enough to significantly affect the apparent gravity vector, but doing that introduces even more uncertainty into whatever number you were to come up with for the gravitational pull the Death Star exerts on the
Executor and the overall effect is just a big question mark.
Beyond that, though, the
Executor was part of a fleet which appears to have demonstrated the ability to decelerate at something in excess of 1000 gravities while closing with the Death Star, and this acceleration is reasonably consistent with the demonstrated accelerations of other spacecraft in other scenes, for example when the Rebel fleet 'pulls up' to avoid crashing into the Death Star's protective shield.
We know of course why the Death Star had a fatal fault in the first place.
Justifying the existence of the thermal exhaust port is very, very easy. Assuming that Alderaan is the same size and mass as Earth, the Death Star needed to deliver at least 2e32 J with the superlaser shot in order to destroy the planet (given the violence of the destruction event, the energy delivered was likely considerably higher). The sequence of events within the movie suggests that Alderaan was destroyed roughly a day before the Death Star was destroyed over Yavin IV. The Death Star was expected to be able to fire the superlaser at Yavin IV, suggesting that the reactors have had sufficient time to generate the 2e32 J required to destroy an Earth-like planet, which puts the lower bound for the power output of the Death Star reactors in excess of 2e27 W, which is roughly 10 times the total output of the Sun. Even very small inefficiencies are going to generate enormous amounts of waste heat when dealing with power outputs of this magnitude and you need some way to get rid of that waste energy before it can build up enough to endanger the Death Star, thus the existence of the thermal exhaust ports.
Justifying why the port targeted by the Rebellion was protected only as much as it was is also easy. This is a design problem, and if we're being at all reasonable we'll acknowledge that the designers do not have infinite time, infinite resources, and freedom from any other constraints which might limit their ability to mitigate the risks posed by the exhaust port. The project costs, in both time and resources, need to be managed, and that means that at some point somebody needs to say when enough effort has gone into any one particular portion of the design. We know that the vulnerability is small relative to the overall size of the Death Star, which may make it difficult to find, especially if the Death Star plans are not available and the attackers are instead required to rely upon whatever information their sensors can pick up. We know that some protection was intentionally provided to the exhaust port as it was ray shielded. We know that there is at least incidental protection for the exhaust port in the form of nearby surface emplacements which can fire along the trench that the Rebel pilots used for the approach to the missile launch point, in the form of other emplacements which cover the approaches to the trench, and in the form of the Death Star's fighter complement. We know that exploitation of the vulnerability could result in the destruction of the Death Star but are told that the probability of successfully exploiting the vulnerability is low (the shot is "impossible, even for a computer," and "one in a million"). While we do not have adequate information for a quantitative cost-benefit analysis, the qualitative information appears to suggest that the risk posed by this vulnerability is fairly low.
If we are the Death Star design and construction team, this does not sound like something that we should provide with additional protection at the cost of additional time and resources, especially if we're already running behind schedule and over budget and have the shadow of that paragon of patience, reason, and positive reinforcement known as Darth Vader looming over us. Sure, hindsight tells us that this is not the best decision that has ever been made, but if we want the project to be completed in a reasonable time frame and at a reasonable cost, we have to be willing to settle for reducing risks to levels which we consider acceptable, not for making the risks go away entirely. It's possible that, once proven feasible, the Death Star will cease to be unique, which means that there's at least a remote possibility that a Death Star will one day engage another Death Star. Should we, as the designers of the Death Star, design the defenses of the Death Star under the assumption that the Death Star needs to be able to resist a full-power shot from a Death Star-type superlaser? Probably not. Should we design the defenses of the Death Star to resist an assault by 10,000 Star Destroyers, despite everything we know about the Rebellion suggesting that such a large fleet is well beyond its resources? Again, probably not. When attempting to mitigate risks as part of a design problem, the risks you focus your efforts upon and the amount of effort put into mitigating each risk need to be dependent upon the magnitude of the risk (which is some combination of the likelihood that the vulnerability will be successfully exploited and the degree of damage which successful exploitation of the vulnerability will cause) or you'll end up wasting a lot of effort protecting against events which are extremely unlikely to occur. There's a vulnerability which can result in the destruction of the Death Star but which can only be exploited by a starfighter making a one-in-a-million shot? That might very well be an acceptable risk; the Death Star has, or at least
should have, a considerable fighter complement of its own to help mitigate the risk posed by this vulnerability, the Death Star's turbolasers, while not exactly what I'd call
effective against the attacking Rebel fighters, were also not completely useless despite being optimized, at least in theory, for use against capital ships, and on top of that it's a very low probability event. It's like saying that WWII-era warships should have had better protection in their smokestacks to protect against bombs dropping down the stack and exploding in an engine room; sure, it's
possible, and if it happens it's potentially crippling or fatal, but it's sufficiently unlikely that protecting against it to a greater extent than is already the case is not really worth the cost.