Why do some circuits use bizarre resistor values?

[Dr. Frank]

In Automotive Electronics, every 1/10 of Cent savings counts for resistors. ..
Therefore, cheap thick film technology, 5%, E 12 is preferred.
Using least possible different values and case sizes also saves money, due to lower storage cost.
Each additional magazine of the pick 'n place machine costs extra money.
Therefore, all of our engineers try to avoid special resistors or capacitors and try to reduce the variety, whenever possible. Design reviews support this cost reduction  process.

Anyhow, in some places, even 1% thin film are still required: At A/D inputs, for sensors for temperature, fuel, pressure, etc.
LCDs often need precise R divider ladders for the different phases.
Also LEDs require 1% resistors for exact initial luminosity because up to 20..30 LEDs in a cluster instrument need to give a totally even illumination. The LEDs themselves come sorted for equal forward voltage, and additionally for equal lumonisity classes, therefore each of these LED selections require one special resistor value, 1%.

Therefore, these "odd" components are sometimes required, even in such extremely low cost apps.
Frank

[zapta]

All those logistics issues are the main reason I switched early in my career from mostly hardware to mostly software.

[tszaboo]

In Automotive Electronics, every 1/10 of Cent savings counts for resistors. ..
Therefore, cheap thick film technology, 5%, E 12 is preferred.
Using least possible different values and case sizes also saves money, due to lower storage cost.
Each additional magazine of the pick 'n place machine costs extra money.
Therefore, all of our engineers try to avoid special resistors or capacitors and try to reduce the variety, whenever possible. Design reviews support this cost reduction  process.

The automotive was always a different industry with it's own weirdness. They sold 30.000.000 VW Golf, so every penny saving has a big effect of course.
I have an Agilent PSU nearby, I've checked the serial number, it was *000379. S oyou see, the number of units sold is ~5 magnitude less, making engineering cost 100000 times more. While an automotive electronics engineer (poor bastards) can spend a month to determine the correct resistor tolerance (I know for a fact this happens), if the guy at agilent spends 100.000 times less time on it, he only has 1 second. I think this could be a very good reason.
SeanB is probably on the right track. I usually go for the 1%, 100ppm resistor for common applications, maybe I should reach for the 0.1% instead? Digikey sais it is "only" 20 times more expensive.

[Zero999]

1) Voltage. You may be building a voltage regulator for (say) 13,8V and there's a voltage divider on the output that steps it down to the 1,25V reference. Well, if you use a 100 ohm resistor on one side you will need a 1004 ohm resistor on the other http://diyaudioprojects.com/Technical/Voltage-Regulator/

It's not worth going for 1004 Ohms. The LM317A has a tolerance of 1% (note that's the A suffix, for the plain one it's 4%) and 100R and 1k gives 13.75V which is within 0.3% of 13.8V.