Thanks for these changes! I think they seem reasonable, on a first look.

I was initially going to suggest that we could avoid setting the Makefile's SHELL variable if we just expanded the two places where we use braces to list the two relevant files, test_count and test_count.lock.

Then it occurred to me that if we do set SHELL = bash as you've done, we might instead be able to simplify our primary target's recipe, which has needed an explicit bash -c "..." shell command since commit 4f23722 in PR #3125.

Would you be amenable to trying that in this PR? I can push a small commit to your branch, if you're willing to grant me permission to do that; otherwise, would you be able to try a patch like this?

--- a/t/Makefile
+++ b/t/Makefile
@@ -46,9 +46,9 @@ test-commands : $(TEST_CMDS)
 
 test : test-commands
        $(RM) -r remote test_count{,.lock}
-       @bash -c ". ./testenv.sh && setup && cd t && \
+       @ . ./testenv.sh && setup && cd t && \
                RM_GIT_LFS_TEST_DIR=no $(PROVE) $(PROVE_EXTRA_ARGS) t-*.sh && \
-               shutdown"
+               shutdown

Also. just a note re the CI builds—it appears GitHub Actions have upgraded their default runners to Ubuntu 24.04, and we'll have to adjust one of our build scripts for that change, so for now, just ignore those CI failures. I'll get a patch up to deal with the CI breakage shortly.