service

I've now put in a warn with the e.getMessage(), but I've left in a debug with the full stacktrace.
This catch block is a little odd in that it's used as programmatic flow for non 200 HTTP responses. I notice that the new API server is sending back a 400 for bad input (including for non-existent query fields and wrong types for non-string query fields), which is invoking the warning triangle which puts the HTTP error message into the tooltip so this block is now getting more use!

The reason for wanting a full stacktrace in debug mode is that both the API and the changed client are new, so any errors in the next few months will be more quickly diagnosed if we can allow the user to easily obtain a stacktrace of the exception.

should be Console.warn

Need to add logic that either 1) modifies config if an existing value of UNIPROT_DOMAIN is found that is legacy.uniprot.org or 2) uses UNIPROT_2022_DOMAIN as a config value so old versions of Jalview still work (up until legacy.uniprot.org goes away).

this comment block should really be after the imports ?

makes sense. I do wonder if there needs to be some indication that any results currently shown are stale, however ?

Added those in. I'm still converting to String with getCanonicalName in the getLogger(Class ...) method because, e.g.
JLogger jlogger1 = JLogger.getLogger(Jalview.class);
JLogger jlogger2 = JLogger.getLogger("jalview.bin.Jalview");
should really return an existing logger in the registry if it exists, and the String name is the registry key.

I was thinking that JLogger did this comparison anyway, but I've followed it through and it didn't. So I've added

if (loglevel.compareTo(this.level) < 0) { return false; }

to JLogger.println so that a comparison is made between the (enums) of JLogger.level and the message LogLevel, and if it's not going to be logged then it doesn't call the actual logger at all (in this case Log4j, so the log4j object is never even touched [even with logger.isDebugEnabled() because the answer should be no]).
This now applies to all log calls.
The check whether Cache.log != null is equivalent to Cache.log.loggerExists() which already happens within JLoggerLog4j.
There is admittedly the wrapper of the JLogger getting brought to forefront of memory, but that's presumably smaller than the old Cache.log which was a Log4j.Logger object.

AWESOME JOB !!!! one tiny niggle I had was that Cache.debug/Cache.error doesn't read nicely, so I've abstracted the logging stuff even farther to a new jalview.bin.Console class: see https://source.jalview.org/crucible/changelog/~br=update%402fJAL-3949_jalview_bin_Console_to_cure_sanity_even_more/jalview/

the reason for conditionals is that Java won't perform the string append operation if logLevel>debug. It saves a minor amount of processing time which is important for code that is executed lots.

You could have .getLogger(Class ..) rather than having to call getCanonicalName ?