Apart from my first blue-rail train set 171, I have always looked at the parts in a train set before deciding to buy.
The parts trigger has increased as my LEGO collection has grown. I think the decision to buy has varied for each set and has varied depending on whether I wanted 1 or many of each set.
From age 9 to a TFOL I found the 12V range very desirable, but due to lack of funds at the time I bought bare motors rather than the train sets. In 1996, just as I was buying up the last 12V spares, my trains went wider than 6 and have stayed there ever since. I ran 10 12V motors in my own 8mm:1ft scale trains till 1 failed.
Then in 2001, with 12V motors being unavailable, I had to convert to 9V, replacing motors 2 for 1 in most locos. I still bought mostly bare motors.
Of the 9V sets, here's why I bought the ones I did (note personal opinions, not pejorative, each person has their own motives):
4525 rail repair digger had finger hinges, a rail worker and suspension wheels (when I didn't buy those from City sets). Bought 5
4537 tankers made a handy rake of wagons. Bought 6
4544 car transporter was a good basic 4-wheel wagon chassis and a last chance to get the uniquely-thin signal gantry pieces used in the 12V range. Bought 6
4547 club car was a good carriage but not a "buy many" set as I didn't have the Metroliner train set to go with it. That is the problem with a derivative extra carriage - you have to have the base set in order to buy many.. This may have crossed TLG's mind when refusing to make an extra carriage set for Emerald Night.
Station 4554 had ramps and looked better in yellow than its red equivalent - I would not have bought it in red, given similar availability where I live. This was better than any of the other stations too - I didn't buy any others.
Cargo terminal 4555 had useful wagons and slider parts that I used in a larger crane in 1996, so this was mostly about parts: high yield = more sales. These might have been the first 9V sets I bought, with enough track (10 straights) to test whether the 12V wheels would work with 9V track. Bought 5
Train set 4559 was not a "buy new" set because it was weird. Yellow bases not useful. I got one second-hand.
Train set 4561 was better and the windows were versatile considering the demise of proper train windows. Black bases were a more useful colour. I bought 2 new and 1 second hand. Also no need to buy separate controllers.
Goods train set 4565 had some interesting parts but it was not a "buy many" set.
Motor 5300 / 10153. Bought at least 50. Most in use and a few spares in case of failure (lesson learned from 12V failure). Some of my trains use 4.
10013 Caboose: bought 10
10014 Log Wagon: bought 10. Logs are one type of load that would make a good train, maybe 3 bought wagons to 1 built wagon?
10015 Passenger Coach: was too bespoke so I didn't buy any. Would have bought some if it had had a mix of wide and narrow windows. Would have made a great conversion to Class 101 DMU as "Daisy" from the Thomas stories.
10016 Tanker. Bought 5, converted to 4-wheel wagons because the 8-wide tank works well with my wider trains.
10017 Hopper. Bought 1. Prototype for my Railtrack hoppers
- I used a similar means of opening the hopper doors.
The My Own Train loco concept was flawed because the design of the train itself was not what I wanted to buy.
10020 Santa Fe Loco: Bought 2, to run as a pair.
10025: bought 5, enough to make a train for the locos. Didn't buy 10022.
10027 Shed: Arches not wide enough for my bigger trains. I took the concept lessons and added them to those of the shed in the 7777 ideas book, but didn't buy the set.
10133 BNSF loco: bought 2, as education in US trains (you can never have just 1 with US trains!), and to compare with Cyclopedia books. Might make a larger version sometime.
10173 Christmas Train: Bought 2. Fun set, has proper train windows. I made a better 6-wide tree wagon from the 2 sets, more like a British prototype log wagon.
9V Track: lots of all types but no 90-degree crossings.
Feed wires: lots, all allocated for experiments or permanent use. Plans for sectional operation, using insulation tape at some track joints to make the sections. My layout currently has 4 sections, each using 2 feed wires. The inner circuit will be similar.
RC trains: Bought no sets. Train set 7897 was atrocious, with its nose being the 2nd most useless part ever.
Bought just a few motors 8866 to experiment with. Torque hopelessly inadequate. This would have been the demise of LEGO trains if we AFOLs had not stepped in to push for a more powerful PF train motor.
7938 Passenger Train: Good stand-alone set. Versatile enough, learnt the lessons after the poor RC train. Having bought 1 I thought I could make a second through-carriage from the parts of the two end ones, so I bought another. I now have a 5-car multiple unit as a basis for 6-wide experiments, to supplement and feed the wider trains.
I haven't bought any of the goods trains yet, because I still have 9V track and plenty of LEGO parts to make any conceivable plant to go with the trains. Now that I'm into scale modelling I'm more likely to make a real-looking hopper system from Technic parts and hence not need the bespoke parts from the good trains sets. So in a sense I've grown out of the play opportunities presented by these sets.
10194 Emerald Night: Bought 1 for experimenting, then another 4 to support my Future Large Steam Engine project. Best steam engine TLG has ever made, so I had to encourage them by buying it. I hope its replacement is just as good, has black wheels (red ones are useless to me) and is not long in coming.
10219 Maersk Train: Not a bad set but not quite what I wanted. I already have larger containers and a pocket wagon
7867 Flexi Track: I'm getting into flexi-track slowly, pending good results from my experiments
with it. If it continues to do well then I should buy some more, maybe up to another 1000 pieces over the next 10 years.
7996 Double Crossover. I was glad to be able to operate this pneumatically. Bought 2
88002 Motors: They performed well in torque testing. I expect to buy more as PF trains grow on me. I will retain the 9V capability for the main line for the time being and run PF in the yard, but a wide flexi-track curve will be tested on the main line in due course. If it performs well then my transition to PF will accelerate.
My experiments with 6-wide sets have supported my wider trains. If the parts yield from a set, for use with my wider trains, is poor then I won't buy.
The new anti-swallow couplings are expensive parts (high proportion of the cost of the sets) and I need only so many of them before I won't buy the sets any more. If TLG separate the buffers from the coupling mount then I could buy the sets freely and widen the buffers.
Wagon baseplates are another high proportion of the cost of a set so these have to be a sensible colour. Black, grey, dark grey OK but white or yellow are silly, resulting in no sale.
I think PF-compatible is a good way to go for locos. Keep the motor etc... separate so that the train set can be a better parts pack or the customer can define their own motorising solution if they wish.
As time has gone by, growing up and getting a job, my LEGO funds have exceeded my ambition to buy the sets and I have become more choosy. I used to have a train set. Now I have a scale model railway with proper scenery. I buy a lot of PaB and fewer sets. Everything's a parts pack and a set has to be spectacular and reasonably priced for me to buy it just for the set it is. What I want from a train set is the right bespoke train parts, good value and a model that gives me an idea, especially if I can convert the set into something for my wider trains. The extra wagon packs were good for that.
LEGO is what it always has been - a new toy every day. The driver is ideas, followed by the parts to make them happen.