You can do this in portions. Create a Base file. Personally would not use an embedded DB but if doing so use Firebird. Define your table. Do not use numeric or decimal field types in the definition. Instead use type DOUBLE. There are bugs with the copy/paste using these types. Now you can copy portions of your data from Calc, starting small to see that all is correct, and pasting into the table - append data.
Note that data is not automatically saved in Base using Firebird until you perform a save on the Base file. Between your copy/paste actions, you should copy your Base file as a backup. Then you have a starting point if a section has a problem.
If opening your .xlsx file in Calc, before transferring data to Base, save it as a native Calc document (.ods format). This may be part of your problem. Only a safe guard - work with native formats…
Delay in replying to additional was due to testing.
Unfortunately, too often only after an answer is posted do the requirements become known. Even then they are usually vague and still probably incomplete. Such is the case here.
Based upon the last comments (posted as answer incorrectly), it is not apparent OP will accept results here or move on to other possibilities. Regardless this post may help others and resolved some of my curiosity. Some results are surprising.
Tests were done on Linux Ubuntu 18.04 Mate (MySQL test on Mint 18.3), LO v220.127.116.11, AMD 4 core 4GB system. The csv file used was 13M and was opened in Calc resulting in a 5M file. There were 81,832 records (21 columns) with mixed data types.
OP multiple times stated Base would hang, but this is most likely not true - only appears so.
First test - copy from Calc & paste to Firebird embedded.
First run all text fields abt 45 minutes
CPU - one of 4 cores varied from 2 to 100% usage
Other 3 cores occasionally used 3-10%
Base appears hung but it is not - it is processing but there is no indicator showing this or any progress type display.
Base screens may even become transparent(except for frame) or display info from other sources while processing.
ALL LO operations are halted (including Calc, Writer, etc.) until process is complete.
Second test modified some fields to reflect DOUBLE vs TEXT types but results were nearly the same as the first test.
Third test - Used macro to read csv and create table. Results were nearly the same with the exception that you could process other LO files albeit slower processing.
Fourth test - On Mint tried MySQL (Workbench) 3+ Min but failed each time & no specifics were presented as to what was wrong.
Back to Ubuntu - PostgreSQL pgAdmin III imported file in less than 30 sec.
Last test - Calc to Base (copy/paste) connected to PostgreSQL via SDBC: just under 12 minutes.
In the tests using macros or Admin tools, care is needed in the csv content. Data in fields may contain quotes, commas or apostrophes which will cause problems. This was NOT a problem in the copy/paste method and thus actually saved some time.
Back to OP’s latest comments, suggestion of importing data in portions was made to help resolve issue of Base “hanging”. Original question never stated this was a recurring process.
It is also not stated what is meant by no longer being portable. Using a DB server doesn’t mean it is located remotely. This is best dealt with as a separate issue.
Could not duplicate the HSQLDB error - not enough info to determine what is the cause.
While there are deficiencies in the process, the final results work. It is unfortunate to see unnecessary criticism of a free product just because it doesn’t work just as an individual wishes. All of us are free to use something else.