Quoting from the Java Docs:

Unfortunately, as the Earth rotates the length of the day varies. In
addition, over time the average length of the day is getting longer as
the Earth slows.

I am building a new planet and I was wondering exactly this, should the day lengths be fixed or should they vary as my planet rotates?

I do not care about the developers in my planet, are there any reasons to say "Unfortunately" before the sentence "as the .. rotates the length of the day varies."? (I have not decided the name of my planet yet, hence the "...".

Actually, I'd argue the opposite; it would be unfortunate in the long term (in geological scales) for life if the length of a day wasn't lengthening.

The primary reason why the length of the day is increasing is because of Tidal Friction, which is caused by the moon. The moon is also getting gradually further away from the earth, and will one day leave Earth's orbit. That's not anticipated to happen before the earth becomes uninhabitable due to the sun increasing in intensity and size, but if it was, it would be a very bad thing for life on earth.

For one thing, the orbit of the moon generates tides and other changes which (aside from seasonal factors) create diversity in the ecosystem that certain animals take advantage of, or at least work to the cycles of. Also, it's thought that the Earth would wobble more without the moon holding it in place, so to speak. The earth already wobbles a little on its axis, causing the Sahara to transition between food basket and desert every few tens of thousands of years, but without the moon it's thought that the earth could eventually drift into a 3 dimensional rotation, creating all sorts of havoc on the surface to different biomes.

In short, the moon does a lot to stabilise the earth while introducing cycles that occur more frequently than seasons. It does that however at a price; tidal friction.

So, it's arguable that most planets with an environment stable enough to form intelligent life, life that can determine its relative fortune according to the environment, probably have a moon. That moon (in turn) is gently robbing the planet of angular momentum, assuming a large body of water. In such a case, intelligent life everywhere is probably experiencing longer and longer days.

I'd argue the best solution is a change in the Java Docs; Unfortunately should be changed to Fortunately.

It really bugs them

What you've probably misunderstood is what the term "Java developer" actually denotes. Java developers are a small mammal similar to civet cats in Indonesia both in physical characteristics and how they are used by humans. They remove bugs from and consume carefully managed Java beans, passing out the undigested remainder into the soil to become source-trees, which is how their common appellation is derived.

As a species Java developers are creatures of habit that are extremely easily disturbed, highly sensitive to changes in their (runtime) environment and typically exhibit signs of OCD. They are so dependent on routine that merely adding or subtracting milliseconds to daytime(and by logical extension, working hours) is enough to cause a marked decline in morale and actual output. This is why in captivity you'll often see them living in production frameworks where their view is carefully modeled and controlled.

Adding a second every 1.5 years or so puts their feeble circulatory systems and nearly nonexistent tolerance under immense stress and is predicted to cause a disastrous decline in their population. This is why it is quietly labelled as "unfortunate" in reference material written by the people who handle Java developers

It is unfortunate, because we don't get to stare at a giant moon all day

If the planet had a moon, and it had a fixed day length, that would mean that it is already tidally locked to its moon. Our moon is tidally locked to Earth, but the reverse is not true. If Earth were also tidally locked to the moon, then day length would no longer change, and one side of the Earth would always be able to see the moon.

Furthermore, in order for two objects to tidally lock each other before the sun turns into a red giant and expands to consume them, they need to be sort of equal in size. That is why Pluto and Charon are locked to each other, but Earth and the moon are not. So in addition to seeing a moon in the sky all day, you would see a HUGE moon in the sky all day. That would be awesome!

Unfortunately, the other side of the Earth would not see the moon at all. Not everyone gets to be a winner.

It doesn't matter

Throughout the vast, vast, vast majority of human history, we neither cared nor noticed that day length varied in the context you're describing.

We cared that it varied throughout a year (i.e., depending on your latitude, the length of the day was longer in the summer than in the winter), and it could be said that it's unfortunate that at the poles day length tends to vary on the order of months, but again, that appears to be outside the context of your question.

OK, maybe it matters

For the vast, vast, vast majority of humans today we still don't care. Well, we care a little bit. We care enough to implement annoying things like Daylight Savings Time.¹ But it should be noted that we were somewhat forced into caring by, among other things, train schedules. Which want to be predictable (even immutable) despite the later advent of busses that can't keep to the schedule to save their proverbial souls in heaven. And the problem with that predictability is the planet's annoying habit of slowing down, perceptually "adding time" to the schedule and apparently requiring all trains to freeze in position somewhere around midnight for just a moment to let the predictable schedules catch up to reality. (Maybe that's why the busses are always late... as a city cost-saving feature they don't freeze at midnight....)²

Which seems to suggest that things in motion tend to stay in motion until acted upon by Man's inability to account for all of the infinite variables that affect the precise measurement of time.³

Rats... apparently it matters

Now, honestly, all of this wouldn't matter at all were it not becoming more apparent that somehow I'm being left behind in the Information Age because my bedside clock cannot talk to my refridgerator to obtain the correct time.⁴ Despite advancing confidently from sundials and water clocks (which were more than enough for my honorable ancestors) to digital clocks,⁵ we humans have determined that it isn't enough. I mean, after all, navigators during the Age of Sail could always use a more accurate clock to figure out where they were on the open sea.⁶ So, apparently, we do need to care about just how long a day is.

Stop babbling and tell me why!

Back in the days when our day started sometime after the sun came up and ended eventually after the sun went down (or after we drank enough mead...) and spent most of our time gathering food and having the ability to actually move when our neighbors got too close to us, it didn't matter at all that the length of a day was predictably unpredictable.

But as we evolved into a species that trained its young in advanced planetary zombie/alien defense (which depends on the mod you're using), allowing for friendships to flourish across oceans, we discovered that it's actually a bit of a trick for everyone to stop doing what they're doing at midnight and hold their breath the requisite time to ensure the next day was really 24 hours long. ...ish. The more we humans must synchronize and coordinate our time, the more important knowing just how long a day is, is.⁷

So, whether it's ocean navigation, train schedules or the latest incarnation of Half-Life, it's unfortunately true that more than just Java developers have to and do care about the measurement of a day's length.

Because time waits for no teenager⁸

_{¹ Thanks to the movie National Treasure, it is popularly believed in the U.S. that Benjamin Franklin is the father of Daylight Savings Time. What he actually proposed in an anonymous letter to a French newspaper was to use cannon to wake the Parisians at sunrise as a method of saving candles — infering, perhaps with reasonable credulity, that the French really enjoy their nightlife. Whether or not this letter constitutes a rational proposal for re-aligning the calendaring process during the course of a year to better distribute available photons for public use is still being debated.}

_{² It's a really good thing the Earth isn't speeding up because asking a train to freeze is a LOT simpler than asking it to momentarily accelerate to the speed of light to compensate for a shorter day....}

_{³ My sincere thanks to Sir Isaac Newton for not climbing out of his grave to slap me with his dislocated left arm for so abusing his otherwise excellent First Law.}

_{⁴ I've had that clock for years and it kept perfectly good time for all that time before my fridge could contact a server farm in India to schedule a time for its own repairs. But now, for some reason, it no longer keeps accurate enough time. Stupid fridge. Time to upgrade my timepiece....}

_{⁵ A 26"x44" version being on sale right now for only $350 USD! The pretty girl looking coyly at me from the depths of my monitor was the final decision maker for me. Oddly, it's lacking the wireless network connection my refridgerator needs to ensure it keeps the correct time. Apparently the folks at Big Time Clocks aren't Java developers.}

_{⁶ And they didn't apparently care when the sun rose or set (other than as a means of avoiding cannon fire). They cared about noon — because if you don't know exactly when noon be, one beint unawarez of where Tortuga and all da rum be! Aaaargh!}

_{⁷ I've waited my whole life to use a double-is like that... really!}

_{⁸ A reasonable PhD thesis could be written on the number of people each individual in the U.S. depends on during the course of a single day — each doing his or her job in a reasonable period of time. And that requires us to care about the precise length of a day to such a degree that it causes concern among Java developers. Think about the precision we'd need if people had to actually do their jobs to the second! It's enough to make a civet's Java developer's head explode just thinking about it. I loved NullPointer's answer, BTW. I upvoted it and you should, too.}

I'm sure you've seen the recent news about the redefinition of the value of the kilogram. Consider the impact this has on every manufacturer that produces measuring equipment: they now have to retool their production lines to use this new definition. Consider the impact this has on every manufacturer that measures things: they have to buy new equipment that uses the new definition, or upgrade old equipment. And that's not even beginning to take into account the systems that have made the assumption that the mass of a KG is the mass of the cylinder stored in France, and the systems that depend on those!

Now consider the impact if the same thing had to happen every one-and-a-half years, because that is essentially what happens with timekeeping systems. As per the same Javadoc:

The actual length of any given day and the amount by which the Earth is slowing are not predictable and can only be determined by measurement.

Because it's impossible to create an algorithm to predict when the Earth's day will extend and by how much, the only way to account for this is to manually modify timekeeping systems to handle each special case. So the system has to say "from 2018-01-08 at 19:48, there are 86400.01 seconds in a day, not 86400.00". Then about 1.5 years later, the system has to have a new condition added that says from that date and time, there are 86400.02 seconds in a day. Repeat ad infinitum.

Modifying software is a tricky business because it's easy to break the current functionality. There are ways and means of alleviating this, but no way to guarantee that a system is 100% correct. This is problematic because millions of systems around the world rely on these timekeeping systems in order to operate; an error introduced by a timekeeping system would therefore cause havoc in those dependant systems.

And then, of course, you have poorly-written software that relies on timekeeping systems always returning the same value. I can guarantee you that probably 90% of all software would break if it asked the underlying timekeeping system for the number of seconds in a day, and received a value other than 86400. Human nature being human nature, the timekeeping system will likely be blamed first as opposed to the incorrect software, which creates more headaches for the people responsible for keeping the timekeeping system correct.