TL;DR
The bottom line is migrations, in either direction, are relatively rare, and consequently don't have an impact on the bottom line. Furthermore, a net migration of questions from EL&U to ELL was always expected, and intended, and will remain the pattern as ELL matures¹.
The plural of anecdote
That said, empirical analysis is always nice, even if it's only to set our minds at ease. With that in mind, I threw together this SEDE query.
Here are the results, broken down by quarter, of questions migrated from here to ELL and vice-versa.
The first thing you'll notice is that your instincts are correct: there indeed has been a recent trend of net migration away from ELU to ELL. But there are two important caveats.
Peanuts
For the first, let's look at the data tabularly, as it's hard to get a solid, immediate sense of scale from the chart. I'll only post the last 3 years this way; the table gets too long otherwise, and data before that is less relevant, which we'll come back to in the second caveat.
Focus on the last column, the number of questions we "lost" to ELL.
+---------------+---------+--------+-------------+
|CalendarQuarter|Departing|Arriving|NetDepartures|
+---------------+---------+--------+-------------+
|FY18Q4 |35 |1 |34 |
+---------------+---------+--------+-------------+
|FY18Q3 |22 |0 |22 |
+---------------+---------+--------+-------------+
|FY18Q1 |13 |0 |13 |
+---------------+---------+--------+-------------+
|FY17Q4 |10 |0 |10 |
+---------------+---------+--------+-------------+
|FY17Q3 |9 |0 |9 |
+---------------+---------+--------+-------------+
|FY17Q2 |2 |4 |-2 |
+---------------+---------+--------+-------------+
|FY17Q1 |10 |0 |10 |
+---------------+---------+--------+-------------+
|FY16Q4 |12 |2 |10 |
+---------------+---------+--------+-------------+
|FY16Q3 |14 |1 |13 |
+---------------+---------+--------+-------------+
|FY16Q2 |6 |3 |3 |
+---------------+---------+--------+-------------+
|FY16Q1 |6 |0 |6 |
+---------------+---------+--------+-------------+
Note that the order of magnitude is 10. Over the course of a quarter; that's 90 days. To put that in perspective, EL&U gets an average of 54 questions per day. Even the highest "loss" quarter, the one we're in, we only migrated 34 questions to ELL. That's about one every 3 days, or about 0.7% of all the questions.
That's not a lot.
Peanut Gallery
But, you reasonably ask, isn't it alarming that the trend is indeed increasing? Ah, this brings us to our second caveat.
It was always envisioned to work this way. Before ELL was set up, the EL&U community realized it had a problem: trying to serve two separate audiences with two distinct sets of needs. We had questions from the nominal audience (linguists, etymologists, and English-language enthusiasts), and even more questions from another audience (people learning English as a second language).
They didn't mix well; it made the people in the first audience grumpy, and people in the second audience frustrated. Thus, ELL was born.
But it takes time to start, build, and establish a site. Rome wasn't built in a day. So while ELL was cutting its teeth, ELU still received dozens of questions a day from the audience we hadn't yet totally migrated to ELL. Now that ELL is mature, more of that is happening.
The hope is, over time, Google results and other organic forces will eventually result in ELL-style questions being asked on ELL from the get-go (and, likely, we're going to have to make some additional, less organic adjustments¹).
But until that Halcyon day, we'll continue to receive these questions, and continue migrate them. That's all. This is to be expected².
Silver Lining
However, if it still worries you, maybe you can be consoled by another fact. If you remove the WHERE mSite=N'ell'
from the SQL query above -- that is, count up migrations to and from the entire network -- over the last 3 years, we've been gaining about 8 questions every quarter.
¹ However, organic changes as ELL matures will likely not be sufficient to totally shift the ELL-asking population to asking on ELL in the first instance. This is because there are several strong signals to Google and other entry points to StackExchange that EL&U is the right place to ask these questions.
These include the body of such questions already existing here -- ironically, to fix that, we need to increase the number of migrations we make to ELL -- but also things like our canonical URL being english.stackexchange.com.
Having said all that, all this goes to decreasing the number of "ELL-style" questions on the site. But that does not, of course, help us increase the number of EL&U questions. We need a seperate remedy for that.
² Contrariwise, it's not to be expected than an ELL user, intentionally posting on ELL, frequently asks a question that requires the expertise of ELU.
And if and when that happens, I wouldn't begrudge the ELL community keeping it, should they decide to do that. The answering community there are also experts in English, and I'm sure they like a meaty question now and again to sink their teeth into.