Let's see if I can just pick out a couple here.
In the early 80's I worked at a place manufacturing avionics systems that had one computer in the plant, an HP3000. The head of the test department had seen me and one of the engineers playing around with our brand new Commodore 64s at the plant on Saturdays when we'd come in to monitor some testing and bring our computers to make the hours between data sets pass faster. He was very interested in what we were doing with using them as programmable controllers, we had some homebrew optocoupler interfaces we'd built that we'd use to do all sorts of silly things. He wanted to built a new automated test rig for our equipment. We had one rig already, run by lawn sprinkler controllers. It wasn't unusual for someone to miss one of the little keys used on the timer dials for these and mess up an entire test run, however. The lawn sprinkler controller unit had replaced an older system controlled with wire-wrapped relay logic boards that had become extremely unreliable.
The chief of test was very interested in using a microcomputer to control the test equipment, since there was no way he'd be able to use the HP3000, which was dominated by MM/3000, an inventory control system, and a number of other management tasks when time could be spared from MM/3000. He was also interested in automated data recording--using the computer to take measurements on the systems under test rather than having a tech with a pencil and clip board read meters and write down the results, then enter them into the HP3000 at a terminal, after which they'd be entered into a report by an RPG program that would print test reports for each unit tested to show that it met its specifications.
Finally he picked up a Compaq portable, practically the day they came out. After playing with it for a few weeks, he hooked it up to the HP3000 as a terminal, so that he could have a terminal in his office, and so that he could try inserting data automatically into the system using Turbo Pascal programs.
Around the same time, the HP3000 had been having a lot of problems. It was seriously overloaded running MM/3000 and an Image database system that they'd started using to maintain some local information drawn from an ISAM database running on IBM mainframes at our corporate headquarters. On top of that, a number of people around the plant were enjoying using it to keep lists of various things for their jobs, it had a simple Listkeeper program that was a sort of proto speadsheet program that was very easy to learn and which several people around the plant were using as a sort of database app. Every day at 4pm the system would crash, leaving the MM/3000 database in a "dirty" state when it was brought back up. On top of that, we would often have crashes at other times of the day, 10am and 2pm being the favorites.
When folks from HP arrived to help us with their problems, they worked for several weeks but we were still having crashes on the 3000. Then they decided to walk around the plant. One of them saw the Compaq, found the leads running into the back of it from the 3000, and announced "Aha! You've got a piece of non-HP hardware hooked up to the 3000! Your service contract is toast!" and all the HP people packed their bags and left. HP told us they refused to support our computer until we removed all non-HP hardware from our system.
I went to our IS (Information Systems, before they started calling it IT) manager and offered to help. I determined that we were short of memory for the tasks we were running, and suggested that they not start backups every day at 4pm, but get someone else to run backups if the IS manager wasn't in a position to stay later for backups (she was a working mother and wanted to get out at a reasonable hour.) The 10am and 2pm crashes were caused by all the leads and line managers in the plant firing up Listkeeper during their workers' break time to update their databases. HP had recommended we buy a completely new system, and supplement it with an HP9000 as a departmental computer for Engineering. We had plenty of processor power for the tasks we were running, however, though we needed a memory upgrade pretty bad. And even though I would really have liked to get a 9000 for Engineering, I didn't see how that would help our problems with MM/3000, the database, and all the Listkeeper users. (In Engineering we were doing most of our computing on HP-41Cs and HP-67s, with a Vic-20, a couple of Commodore 64s, and an Atari 800 squirrelled away in the back corners of some of the offices. Only one engineer regularly did computing on the 3000, and he had a vi clone he'd written in Fortran/3000 that required so much computing power that he was banned from using it during business hours, so he'd come in at 10pm, after his wife had gone to bed, and use it and a Spice clone he'd written on the 3000 for a couple of hours.
Well, my analysis turned out to be worth everything that management had paid for it, from their perspective. They banned the evil Compaq from being connected to the HP3000's serial line. They made amends by purchasing two HP150 computers, one for Test and one for Engineering. They also claimed to have a line on a used HP9000/300 from another division of the corporation that would be provided to our engineers, as soon as the other division could find its software tapes and box it up. Mollified, HP came back on site. I dropped off a copy of my own testing and writeup in a not-too-inconspicuous spot near the 3000 in the hopes that the HP reps might find it useful. A couple of days later, once the HP reps figured out a new system was not in the cards, they announced that we could probably get by for another year if we expanded RAM and didn't do backups during the business day.
A month later, MM/3000 and the database were running smoothly, one of the HP150s had been moved to the production floor where it ran Listkeeper for the line leads, the other ran Listkeeper for Test and Engineering, the Compaq was hooked back up to the 3000 using a camoflaged cable. I was on the 3000, demoing real time graphical data monitoring to management using the HP3000 User Group's version of Forth. When I started the program and had it bouncing bar graphs around on the terminal screen, the production manager actually screamed and ran from the room. He returned a few minutes later to report that my program had not only failed to stop MM in its tracks, but had actually had less impact on the computer than a single Listkeeper session. Thus I introduced plant management to the idea that if you don't buy big fat apps from HP, you can actually do a whole heck of a lot with the computing power we had.
Another time, another tale...
I remember a group of 12 engineers doing fluid dynamics simulations for rocket designs on a shared Dell PC-XT clone while a new 486 sat unused on their manager's desk. The manager had to have the new system, as it was a status symbol. The keyboard was placed in a desk drawer, however, as he didn't want to look like he did clerical work.
I was in another area, where I had a 286 all to myself that had been cobbled together from parts literally pulled from trash cans around the plant. One of the engineers came over to use my 286 regularly, as I was only using it to maintain paperwork on calibration schedules and lists of instrumentation required for different tests. He wondered how I managed to hang on to a 286 for myself when everyone else was fighting for time across the plant on shared XTs and VAX IIs.
My trick (other than the fact that nobody had any paperwork on my system) was that it looked like a pile of electronic garbage, not a computer. It was built up on a partial chassis from a different type of equipment (some old boat anchor, I've forgotten what it was originally) wires were everywhere, the video cable was hand-made from individual wires, the monitor had a case that looked like a truck had run over it then been duct-taped back into shape, the keyboard had been left out in the weather for the better part of a year and looked awful though it worked great.
We hatched a plan for improving things over in his work area. I contacted the manager and offered to put together a 286 for his engineers. He agreed to allow it, so long as he didn't have to pay for it and no other managers showed up wanting to know where it had come from. I gave him a story to tell other managers that it had originally been received as a controller component for another piece of instrumentation (which is why IT had no record of it, and why he'd been allowed to get it) and that the instrumentation it was married to was destroyed in a test, freeing it up for use as a general purpose system (explaining why it wasn't mounted in a rack in the control room.)
I didn't offer to put together a 386 for his engineers because I knew such an offer would cause problems. First, a system so close in power to his own mighty 20MHz 486 might threaten his own ego, especially if he suspected somebody might run it up to 25MHz. Second, he wouldn't be able to keep it. Somebody from elsewhere would find out that mere rocket engineers were using a 386 while some manager elsewhere still had a 286 on his desk, a pissing match would start, and the other manager would get the 386, leaving the 486 manager "beaten" because he couldn't hold on to the equipment. A 286 was enough of a jump for the engineers to make it worth the manager being willing to agree to a new system, but it wouldn't cause any political waves.
Then I pulled together all the parts necessary. I made a copy of the 286 motherboard's BIOS ROM, and edited it to give the same message as the manager's 486 BIOS at startup then burned a new chip and dropped it in the 286 board, along with a label over the top of the 286 which read "Intel 486 20MHz". Then, after hours, I and a couple of other engineers pulled the motherboard out of the 486, edited its ROM to report itself as a 286, and stuck the mobo in a dumpy old 286's case with the I/O and Hercules graphics cards I'd scared up, while putting the 286 board in the manager's machine. Fortunately the manager had very few files on his system other than the ones originally installed by IT in a standard software configuration, so we were able to make the hard disk swap without much trouble (the 486 used IDE, while our 286 host adapter was ST-506.) I moved over everything from 1-2-3 and Harvard Graphics, the only two apps the manager had ever run, on floppies.
The manager was never the wiser, and the engineers got their fluid dynamics done so much faster on the "286" that the manager treated me to lunch because he was so happy that they weren't whining at him about their slow computer any more.