MAINJAFAD, O-P:
As a former USMC avionics tech (A-6E, F/A-18A* FLIR/l a s e r** systems, "I" level), I can say definitively that you both hit the nail right on the head with regards to tolerances.
If "good enough" was accepted to send a LRU out the door, it was guaranteed that the matching LRUs in 30%-40% of the aircraft it could be installed on were also at the edge of tolerances, and the combination would result in an immediate "fault" condition.
The only way to make sure of success was to align everything to as close to "ideal" spec as possible.
Additionally, as Leon Jabachjabicz noted, operating environment had a major effect. At MCAS El Toro (Santa Ana, CA) the air was normally moderately dry, and we were far enough inland to not get salty air (actually, the normal breezes/winds were from the more arid areas further inland), and we had little problem with corrosion.
At MCAS Iwakuni, Japan and NAS Cubi Point, Philippines, however, the air was more moist and hot - leading to corrosion of cable pins, etc. This was even more pronounced aboard ship - to the point that ~70& of the faults that LRUs (line-replacable units) came in with could be cleared by simply unplugging all the internal cabling, spraying cleaner on all the pins & contacts, and re-connecting them. This quickly became our standard "first action", even before hooking it up to the test bench to run a diagnostic.
The problem was that the humidity was allowing a fungal film to grow on the connector pins, which blocked the low-power current flows of digital electronics. This was normally invisible - but if left untreated it would become visible, and even begin pitting the plating on the pins!
MAINJAFAD also has a point on the "design & make" aspect. The AN/ALQ-126 Defensive ECM system on the A-6E (and other aircraft) was a very low MTBF item in the early 1980s-with MTBRs around 10 hours. The LRU was a two-deck box, full of circuit boards that were covered with IC chips. When Sanders developed the -126, they "had a brilliant idea" - to "make maintenance easier", they put almost all of the IC chips in plug-in sockets (the sockets were soldered to the boards). Over 90% of failures were cleared by "simply" reseating all of the >300 IC chips in the LRU - a process which took several hours, but which had to be done, as invariably in each LRU there were multiple chips that had vibrated loose enough to break connection - despite the "hold-down clamps" Sanders had installed to prevent that.
The solution was developed and tested by my A-6E squadron ( VMA(AW)-242 Black Bats at that time) while on deployment to MCAS Iwakuni the summer of 1984 - my good friend Kevin convinced his superiors to let them modify 3 -126 LRUs by removing all of the sockets and soldering the chips directly to the board. The modified LRYs immediately went to over 100 hours between failures, and once NAVAIR approval was granted to modify all the LRUs in the squadron, the MTBF eventually settled at about 200 hours - a 20-fold increase. Troubleshooting and repair time also dropped dramatically, as now there was usually only one point of failure, which was easily and quickly located and replaced.
Sad to say, when the change was made official throughout the Navy/Marine Corps aviation community, the credit was given to the squadron's avionics maintenance warrant officer. He had argued strongly against Kevin's suggestion, but once the test worked well, he had submitted the "Bennie-sugg" (Beneficial Suggestion Program) paperwork with his own name as "originator of suggestion", thus the credit (and financial award) were given to him, and NOT to the ones actually responsible.
* While my squadron ( VMA(AW)-121 Green Knights at that time) was deployed aboard CV-61 Ranger in 1985-1987, I also worked on the S-3A FLIR, and worked beside the guys working on the cameras from the F=14 TARPS recon pod.
** stupid board auto-scramble function - can't even spell l a s e r properly without it changing things... laser.