TL;DR; Bulk, in essence, is just relevant when the items are not inside some inventory.Packed items like pallets/containers
- Just adds another thing to do for every single recipe, mostly just adding tediousness. Basically mandatory barrels if you wanted to increase throughput.
- ...
Stacking should mostly be done as a convenience for the player, not by the player themselves. Every item should be stack-able and consumers should be able to consume/output bulk. Inserting bulk and single of the same item-type is no issue - the assembler would always unpack inputs into its internal slots (when you open the assembler, it would always be a number of single items). As an example, an assembler making inserters would get a bulk of iron plates (for this post, ten iron plates), ten single gearwheels and a bulk of green circuits. It would produce a bulk of inserters, taking ten times as long as the normal recipe. Optionally, make a setting what to output or add a new "bulk-assembler" that waits for enough input to produce a bulk output. Otherwise, just take bulk as soon as one input is bulk. Note that even mixed belts of bulk and single are behaving quite well. Importantly, less items should mean more UPS!
Sure, there would be a way to change a bulk to single and back, but it would rarely be used (and there could be hotkeys for doing it manually, and a general "pack"/"unpack" recipe for assemblers). A large throughput factory would just do bulk all the way. There is another nice benefit: on the current version, megabases with higher mining productivity have drills fill express belts insanely fast. People use bots even for mining, as creating patterns that output and use enough belts for larger patches is tedious. Mining drills should instead be able to output bulk ore at a tenth of the speed. Furnaces would then take bulk and smelt that, and after more steps, bulk of science would finally be taken by science labs.
Most importantly, note that
see this post, it also includes other interesting ideas. This does not nerf robots but increases throughput by belt considerably. Using bots to carry low volume still works as before. Intuitively, a flying robot cannot carry a bulk of steel beams.[...] bulk objects [...] cannot be flown by robots.
In all typical cases, factories do not produce or consume a single of something. As soon as the time for an assembler to be supplied with ten of an item type is irrelevant, bulk will be the option. This is the case even early in the game and for a vast majority of all items produced (think science, who cares if it is ready a few seconds later).
Balance could be solved by making bulk conversion a late game research or by other methods, e.g. rebalancing early game throughputs (yellow/red belt speeds, etc).
This highlights the issue that item price is generally too low and quantity too high. One rarely needs single of anything - not even for personal use and especially not for production. A megabase would rarely ever use single items, in essence just multiplying item price by ten and reducing quantity equally. Note that when stack sizes would always be multiples of bulk, even placing entities from bulk would be no issue, the bulk slot would just automatically be converted to single (potentially needing some time, but no explicit "unpack" command by the user would be needed).
________________________________
Unrelated addition requests (sorted, first being the most important):
- Solar power MK2. Placing solar power is the issue, not making it. There are bases with two million solar panels - two §"$%"§$ million! With nuclear being as UPS hungry as it is, there is no alternative to solar sadly.
- Make nuclear less UPS hungry. It takes fuel cells in, outputs electricity, and is almost never looked at after placement. There is no reason for it to use so much cpu. A related point is that uranium spawns too much - or fuel cells, nuclear fuel and nukes are just super cheap! Capitalizing on the latter, in addition to barely anyone ever building single reactor setups, make all components of nuclear reactors twice as expensive, increase the fuel cell cost by ten (produce one fuel cell instead of ten), and increase heat/steam/power production/consumption by four. Potentially reduce neighbor bonus. This makes reactor setups below four a viable option, reduces cpu used by nuclear by factor four and finally gives me a way to get rid of the uranium on our map - i set it to the minimum where i could notice any uranium spawn, and cannot possibly use it all. A single patch, 2 million ore, was enough for the entire game. It was harvested before researching considerable amounts of mining productivity. As another example, i want to remove an uranium field containing a few millions, but doing so would give me around half a million u-235 (taking mining productivity into consideration). I cannot possibly use that.
- Add bigger, larger chests. Storages use hundreds of storage chests, which should be replaced by bigger versions (why not a 4x4 tile size chest that can hold considerably more than 16 chests?)
- Add a way to load your personal items (roboport, exoskeleton, nightvision, battery, ...) from your base's electrical network
- Add a way to plant trees or generally decoratives that prevent the map from becoming flat and grey (also a good way to use wood! Just make wood be usable to plant trees, this also gives the game option to plant forests between you and biters. Possibly biters need to more efficiently remove forests when they are used as walls)
, I hope this will be read by people who know how to do Russian 
