Thinking outside the box
Most beginning track planners automatically try to center yards and passing tracks on long straight or relatively straight stretches. However, often there isn't room for them this way. It is often useful to wrap them around corner or end curves. This is useful for yards, as the ladder(s) can be on relatively straight sections, making coupling and uncoupling easier and avoiding the need for curved turnouts which may not be readily available.
Aisles and access
Many published layout designs have impractically narrow aisles. 24" aisles are fine for a single-operator layout. However, as soon as you go to two operators, 30" (often shown in plans) isn't really enough. The more common modern advice is 36" aisles at most points on any multi-operator layouts. If you have a high density of operators and/or numerous visitors, 48" aisles at most points may be necessary. These are commonly found on club layouts for this reason.
A double-deck layout will probably have more operators for the same room size, thus it may require wider aisles. In fact, as long as the traffic density remains the same, any measure to increase length of run (such as a twice-around) will increase the number of operators. Likewise, a smaller-scale layout may require more operators. The important factor underlying all of these is the desired traffic frequency.
In John Armstrong's track planning books, he places much emphasis on the "square". However, his descriptions and examples of how to use it show some assumptions and biases.
Squares are probably most useful in HO. This is where they match his description. For example, take an HO layout with 24" ("conventional") curves. Its square will be 28.5" on a side. The equivalent N layout will have 13" curves and a 15.5" square. The equivalent (US 2-rail) O layout will have 44" curves and a 52" square.
On the HO layout, a "blob" requires good access around it. On the N layout, it is just feasible to have a turnback curve accessible from only one side. On the O layout, the center of a blob cannot be reached from any angle, and you have to think much more in terms of curves accessed from the center instead of the outside.
Also, his use of squares assumes that the typical train will be substantially less than the circumference of a minimum mainline radius circle. This emphasis on curve radius instead of train length is (once again) suited to HO or larger. If trains are long (much easier to achieve in N), length of run can become more of an issue than curve radius.
Double-deck layouts have another scale-related issue. Grades are the same in all scales. Deck heights are determined by human ergonomics, which don't change with scale. Thus, in a smaller scale, a train will have to travel more scale distance to reach another deck at the same grade. Looked at another way, this means that the minimum practical size for a double-deck layout is about the same in N and HO.
On a smaller scale, vertical factors may not scale down exactly, either. The thickness of roadbed doesn't vary much with scale. Thus, if two tracks must cross, the increased vertical separation necessary for a tunnel (as opposed to a bridge) will be more scale feet in the smaller scales.
When using or adapting a plan from a book, you must be aware of the assumptions behind the plan. These will not all be explicitly stated, but can often be deduced.
Some of these unstated assumptions can affect curve standards, such as whether you use truck-mounted or body-mounted couplers, or whether or not you have brass steam locomotives.
Plans from the 1950s, on average, look rather different from modern plans. Yards were almost universally single-ended, even the arrival/departure tracks.
Staging was almost unheard of. Its use was beginning in Germany and Switzerland, but the innovation had not crossed the Atlantic. In this time, "ready-to-run" meant "toylike" (it was seen as low-end, which became a self-fulfilling prophecy). Modellers rarely had more equipment than they could fit on their layouts. This will seem very strange to any modern modeller. On the occasions when staging was included in a plan, it was usually termed "storage" or "layover" trackage. The latter term implies an operating scheme where trains usually return from storage, as opposed to only coming out once. Such a scheme, not in use on most layouts today, would have been extravagant to the 1950s modeller.
Gratuitous return loops were common. It was taken for granted that trains would often be turned.
Industries were small; the large multi-track industry had not found general favor.
In that time period, walkaround throttles didn't exist. Thus, a common operating scheme for large layouts was to have several fixed control panels, each controlling an area rather than a train. This was operational layout design, but prioritizing the tower operator's perspective rather than the engineer's perspective. On such layouts, you usually couldn't follow the mainline without backtracking and ducking under... but you were not expected to in normal operation. This style has fallen from favor.
British and European plans almost universally assume passenger service will be a feature, and often that it will be the main feature.
The density of railroads in these countries is much greater than in North America. Mainlines are almost universally double-track or more, and even branches may be double-track. Since British trains run on the left-hand track, this will affect siding arrangements.
Also, especially if the plans are meant to be steam-era, the frequency of 4-wheel freight cars will mean small siding capacities by North American standards. Likewise, they will usually not be designed for long trains. Real European, and especially British, trains are short. Their steam never got nearly as large as in the US, and multi-unit diesel or electric lashups are unusual except in Switzerland. Thus, selective compression of trains is not something given much consideration.
A less obvious feature is that British and European plans generally avoid 90-degree crossings. These are very rare in most of these countries.
Also, advice on layout design (especially as it relates to operations) is often era-specific but not stated. For example, a book may describe how to design an engine terminal, and only discuss how to handle steam.
Relating to "Aisles and access" above, whether you use magnetic uncoupling or a pick can affect the design. With manual uncoupling, a near-eye-level layout becomes impractical.
Curve standards and prototype
Not just your choice of era, but the specific railroad you're modelling, can determine your curve standards.
A common statement is that curve (and overall space) requirements increase as you approach the modern era. This is an oversimplification.
If you're modelling the 19th century, all engines and cars are small. This is the one case where the axiom is true.
As the steam era progresses, locomotives get larger. However, in the first quarter of the 20th century, the most space-demanding equipment is usually passenger rolling stock. In this time period, the principle that freight trains can handle tighter curves than passenger trains is true. It will be true again in one later period.
In the late steam era, locomotives are large. Often, the new power demands curves wider than even passenger cars require. Since, in steam days, there was usually only one engine per train, and larger trains require larger engines, train length now becomes a significant factor in determining curve radius. Thus, branch lines of this era, which have short trains, are in a situation more like the early 20th century. Their engines are smaller, but their passenger equipment isn't.
Early freight diesels were often small. Passenger cab units were longer, though. In the transition era, steam engines will thus determine radii, as diesels will not be the governing equipment.
Immediately after the end of steam, most freight diesels were still small. C-C power was still for low-speed applications. Passenger trains, with A1A-A1A cab units and long cars, require bigger curves. The early 20th century curve situation returns. Train length doesn't matter much, as longer trains are pulled by more, not larger, engines.
But then, as the 1960s progress, C-C freight power becomes very common. Piggyback flats and autoracks over 80' long come into use. Thus, freight trains start to demand broader curves.
Examples of exceptions to the rule in this time period:
If you're modelling a coal branch of, say, Southern or L&N, you can get by with sharp curves. Branches were assigned 4-axle power, long freight cars probably wouldn't see these lines, and branchline passenger trains were mostly gone.
The Western Pacific never owned any 6-axle diesels. Thus, rolling stock - both passenger and freight - wil be the governing equipment on a diesel-era WP layout. However, WP did see frequent runthrough power from BN and UP, and the latter was mostly 6-axle in the 1970s. (Also see next entry.)
UP's mainline requires huge curves. In the 1920s onward, it saw 4-12-2s. These were displaced by Challengers and Big Boys. As steam declined, the gas turbines arrived, and the 8-axle 4500 hp turbines require large curves. The turbines were replaced by UP's legendary 8-axle double diesels. Thus, from the 1920s through the 1970s, curve requirements will stay roughly constant. If you're modelling the early 1980s, you can get by with less, because the only remaining double diesels, the DDA40X "Centennials", were in storage. In 1984-85, the Centennials came out. By this time, UP had bought WP. DDA40Xs became a regular sight on the Overland Mail through the Feather River Canyon. Thus, modelling WP after the UP takeover will require larger curves than while it was independent. After 1985, modelling UP ceases to require larger curves than any other road of the same time.
Compression of train length is not often discussed in British and European sources, probably because their real trains are shorter.
It is very common for passenger trains to be minimally compressed while freights are reduced substantially. This may reflect a preference for passenger service, or simply the fact that a passenger train's cars serve different and interrelated functions while freight cars are independent.
Tony Koester pointed out that a train looks long if you can't see both ends at once. His layout at the time, the Allegheny Midland, featured Appalachian scenery, which was good for breaking up trains. Modelling the plains makes this somewhat more difficult.
Some modellers have observed that a train of the same scale length looks shorter in a smaller scale. This is because it's easier for the eye to take it all in at once.
A train of 20 40' cars looks longer than a train of 10 passenger cars, even though they are the same length. It's easier to judge how many cars there are than how long they are, especially if one is observing the train at a very oblique angle.
Suspension of disbelief
This is an important part of selective compression, and of much else about model railroading (such as Freelancing). You must know your own limits and plan within them. Some modellers, for example, cannot accept staging, hidden cutoff tracks, and the like, because they cannot ignore tracks they know are there.
This is a factor in whether or not you can accept more-than-once-around layouts or other scenic insincerities.