First, there needs to be a distinction between what reality is, and what reality feels like. I ran some tests , and they fall in line with Xisuma's distribution, I used violine's 1655054676 seed and tested from 0,0 to 1023,1023. Not as big of an area, but should produce something hopefully conclusive. (1.16.5 in blue, 1.17 in orange) Even with a decent sized sample, using just my data it looks like it's best to mine at Y=8 in 1.16.5, and Y=12 in 1.17, when that's not exactly the case. To me, this means that it's possible Xisuma didn't get it quite right, possibly. Without a test with hundreds of thousands of blocks instead of just thousands, there's not much certainty here.
[media]
The distribution is fine. Diamonds seem to generate more at Y=14 through Y=16. The amount is also fine. I did run a test with the 1.18 datapack, and there are almost no diamonds above Y=10, you'll have to go below 0 to find a good amount, but the generation is also nerfed on the pack. In a 9x9 chunk area (roughly 150x150), there were a total of 241 diamonds between Y=-64 and Y=-40. Scaling that up to the 1024x1024 test without the datapack, it seems that the number of diamonds distributed between -40 and -64 is roughly the same*?* as between 1 and 16 normally. It's the same amount of diamonds, but the amount is much more spread out, so it's much harder to find diamonds while strip mining with the datapack installed. Some people may have this and not be able to find diamonds, and I know that's not this specific issue but it's worth knowing the difference between the two versions. Part II is still being worked on so I'm sure this won't be final anyways. So as long as you don't have the datapack installed and enabled, ore generation should be fine.
(Y=11 Still seems like the best place to mine, since lava lakes can get in the way below this level. If anything more diamonds at 14-16 should help since lava may cover diamonds below 11 and air pockets around these lakes will expose these higher Y levels.)
With the original sample size of a 100 and 200 block radius, there's going to be some difference between versions, because ores are in fact generated differently. However, there is some randomness in the number of diamonds generated. A difference of 3% even over hundreds of samples seems explainable by chance alone. If, let's say hundreds of thousands of samples are collected, and the percentage increases even more, then it may be fair to say that there is a significant difference in diamond generation between 1.16.5 and 1.17, and there are mathematical ways to test this. Even still, violine got roughly 3.9% more ores on 1.17 in their test, which may also be chalked up to random chance. Only with more samples do we get more confident in the results, it's just a question of where do we reasonably draw the line? Distributions have standard deviations for a reason, not everything is going to be exact.
I know it may not feel right, but diamond distribution is really not all that different. I did run another test, however. I took the blocks in a 2x3 tunnel running for 5000 blocks at Y=11 in both versions of the game on the same seed. I did not write down the exact numbers, but I can tell you the number of diamonds in these tunnels ranged from 23 all the way to 72 on both versions. That's the difference between a full set of tools and a bit of armor, and a full set of tools and armor for you and someone else. Due to the high variability, you may have just gotten either lucky or unlucky. And with millions of people playing the new update and some of them inevitably getting unlucky, they take to the bug tracker and report it as a bug. If you got the 72 diamond tunnel before 1.17 and after updating got the 23 diamond tunnel, you'd obviously suspect something to have gone wrong in the ore generation. Confirmation bias is strong with this one. This is why large-scale testing is important.
[media]
Oh, and here's 1.16.5 and 1.17 diamond generation side-by-side. 1.17 is the top one, with the amethyst geode. They look roughly the same, but with a closer look I think that 1.17 generates ores in smaller veins, (maybe in this sample less frequently too), but I found singular diamond ores seemingly more often, but there's not a way that I know to test this without counting manually. I overlayed a potential strip mining path in red and blue, with one being a worst-case scenario and blue being a best-case scenario. Both of these should be equally likely to happen, so if you get blue in 1.16 and red in 1.17, I can see why one may think diamond generation was changed. (It's crazy that they're only a few blocks off!) However, due to the random nature of ore generation, you're bound to get closer to the average number of diamonds the more you mine, eventually. It's okay to get unlucky sometimes, but statistically, bad luck won't last forever. I hope some of that made sense.
TL;DR - It's all down to luck. On average the distribution seems roughly the same, with a bit more diamonds between Y=14 and Y=16 (This is all true long as you don't have the experimental 1.18 datapack on). Aside from that, it's hard to say. You may get unlucky, but just keep mining!
I'd like to add my two cents to this.
First, there needs to be a distinction between what reality is, and what reality feels like. I ran some tests , and they fall in line with Xisuma's distribution, I used violine's 1655054676 seed and tested from 0,0 to 1023,1023. Not as big of an area, but should produce something hopefully conclusive. (1.16.5 in blue, 1.17 in orange) Even with a decent sized sample, using just my data it looks like it's best to mine at Y=8 in 1.16.5, and Y=12 in 1.17, when that's not exactly the case. To me, this means that it's possible Xisuma didn't get it quite right, possibly. Without a test with hundreds of thousands of blocks instead of just thousands, there's not much certainty here.
[media]The distribution is fine. Diamonds seem to generate more at Y=14 through Y=16. The amount is also fine. I did run a test with the 1.18 datapack, and there are almost no diamonds above Y=10, you'll have to go below 0 to find a good amount, but the generation is also nerfed on the pack. In a 9x9 chunk area (roughly 150x150), there were a total of 241 diamonds between Y=-64 and Y=-40. Scaling that up to the 1024x1024 test without the datapack, it seems that the number of diamonds distributed between -40 and -64 is roughly the same*?* as between 1 and 16 normally. It's the same amount of diamonds, but the amount is much more spread out, so it's much harder to find diamonds while strip mining with the datapack installed. Some people may have this and not be able to find diamonds, and I know that's not this specific issue but it's worth knowing the difference between the two versions. Part II is still being worked on so I'm sure this won't be final anyways. So as long as you don't have the datapack installed and enabled, ore generation should be fine.
(Y=11 Still seems like the best place to mine, since lava lakes can get in the way below this level. If anything more diamonds at 14-16 should help since lava may cover diamonds below 11 and air pockets around these lakes will expose these higher Y levels.)
With the original sample size of a 100 and 200 block radius, there's going to be some difference between versions, because ores are in fact generated differently. However, there is some randomness in the number of diamonds generated. A difference of 3% even over hundreds of samples seems explainable by chance alone. If, let's say hundreds of thousands of samples are collected, and the percentage increases even more, then it may be fair to say that there is a significant difference in diamond generation between 1.16.5 and 1.17, and there are mathematical ways to test this. Even still, violine got roughly 3.9% more ores on 1.17 in their test, which may also be chalked up to random chance. Only with more samples do we get more confident in the results, it's just a question of where do we reasonably draw the line? Distributions have standard deviations for a reason, not everything is going to be exact.
I know it may not feel right, but diamond distribution is really not all that different. I did run another test, however. I took the blocks in a 2x3 tunnel running for 5000 blocks at Y=11 in both versions of the game on the same seed. I did not write down the exact numbers, but I can tell you the number of diamonds in these tunnels ranged from 23 all the way to 72 on both versions. That's the difference between a full set of tools and a bit of armor, and a full set of tools and armor for you and someone else. Due to the high variability, you may have just gotten either lucky or unlucky. And with millions of people playing the new update and some of them inevitably getting unlucky, they take to the bug tracker and report it as a bug. If you got the 72 diamond tunnel before 1.17 and after updating got the 23 diamond tunnel, you'd obviously suspect something to have gone wrong in the ore generation. Confirmation bias is strong with this one. This is why large-scale testing is important.
[media]Oh, and here's 1.16.5 and 1.17 diamond generation side-by-side. 1.17 is the top one, with the amethyst geode. They look roughly the same, but with a closer look I think that 1.17 generates ores in smaller veins, (maybe in this sample less frequently too), but I found singular diamond ores seemingly more often, but there's not a way that I know to test this without counting manually. I overlayed a potential strip mining path in red and blue, with one being a worst-case scenario and blue being a best-case scenario. Both of these should be equally likely to happen, so if you get blue in 1.16 and red in 1.17, I can see why one may think diamond generation was changed. (It's crazy that they're only a few blocks off!) However, due to the random nature of ore generation, you're bound to get closer to the average number of diamonds the more you mine, eventually. It's okay to get unlucky sometimes, but statistically, bad luck won't last forever. I hope some of that made sense.
TL;DR - It's all down to luck. On average the distribution seems roughly the same, with a bit more diamonds between Y=14 and Y=16 (This is all true long as you don't have the experimental 1.18 datapack on). Aside from that, it's hard to say. You may get unlucky, but just keep mining!