diff --git a/.aspect/bazelrc/convenience.bazelrc b/.aspect/bazelrc/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/.aspect/bazelrc/convenience.bazelrc
+++ b/.aspect/bazelrc/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/.aspect/bazelrc/correctness.bazelrc b/.aspect/bazelrc/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/.aspect/bazelrc/correctness.bazelrc
+++ b/.aspect/bazelrc/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/.aspect/bazelrc/performance.bazelrc b/.aspect/bazelrc/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/.aspect/bazelrc/performance.bazelrc
+++ b/.aspect/bazelrc/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links
diff --git a/e2e/bzlmod/.aspect/bazelrc/convenience.bazelrc b/e2e/bzlmod/.aspect/bazelrc/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/e2e/bzlmod/.aspect/bazelrc/convenience.bazelrc
+++ b/e2e/bzlmod/.aspect/bazelrc/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/e2e/bzlmod/.aspect/bazelrc/correctness.bazelrc b/e2e/bzlmod/.aspect/bazelrc/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/e2e/bzlmod/.aspect/bazelrc/correctness.bazelrc
+++ b/e2e/bzlmod/.aspect/bazelrc/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/e2e/bzlmod/.aspect/bazelrc/performance.bazelrc b/e2e/bzlmod/.aspect/bazelrc/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/e2e/bzlmod/.aspect/bazelrc/performance.bazelrc
+++ b/e2e/bzlmod/.aspect/bazelrc/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links
diff --git a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/convenience.bazelrc b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/convenience.bazelrc
+++ b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/correctness.bazelrc b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/correctness.bazelrc
+++ b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/performance.bazelrc b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/performance.bazelrc
+++ b/e2e/bzlmod_write_source_files_external/.aspect/bazelrc/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links
diff --git a/e2e/copy_to_directory/.aspect/bazelrc/convenience.bazelrc b/e2e/copy_to_directory/.aspect/bazelrc/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/e2e/copy_to_directory/.aspect/bazelrc/convenience.bazelrc
+++ b/e2e/copy_to_directory/.aspect/bazelrc/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/e2e/copy_to_directory/.aspect/bazelrc/correctness.bazelrc b/e2e/copy_to_directory/.aspect/bazelrc/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/e2e/copy_to_directory/.aspect/bazelrc/correctness.bazelrc
+++ b/e2e/copy_to_directory/.aspect/bazelrc/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/e2e/copy_to_directory/.aspect/bazelrc/performance.bazelrc b/e2e/copy_to_directory/.aspect/bazelrc/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/e2e/copy_to_directory/.aspect/bazelrc/performance.bazelrc
+++ b/e2e/copy_to_directory/.aspect/bazelrc/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links
diff --git a/e2e/workspace/.aspect/bazelrc/convenience.bazelrc b/e2e/workspace/.aspect/bazelrc/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/e2e/workspace/.aspect/bazelrc/convenience.bazelrc
+++ b/e2e/workspace/.aspect/bazelrc/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/e2e/workspace/.aspect/bazelrc/correctness.bazelrc b/e2e/workspace/.aspect/bazelrc/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/e2e/workspace/.aspect/bazelrc/correctness.bazelrc
+++ b/e2e/workspace/.aspect/bazelrc/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/e2e/workspace/.aspect/bazelrc/performance.bazelrc b/e2e/workspace/.aspect/bazelrc/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/e2e/workspace/.aspect/bazelrc/performance.bazelrc
+++ b/e2e/workspace/.aspect/bazelrc/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links
diff --git a/lib/tests/bazelrc_presets/all/convenience.bazelrc b/lib/tests/bazelrc_presets/all/convenience.bazelrc
index 3122d3273..cb02f1dfd 100644
--- a/lib/tests/bazelrc_presets/all/convenience.bazelrc
+++ b/lib/tests/bazelrc_presets/all/convenience.bazelrc
@@ -1,7 +1,6 @@
 # Attempt to build & test every target whose prerequisites were successfully built.
 # Docs: https://bazel.build/docs/user-manual#keep-going
 build --keep_going
-test  --keep_going
 
 # Output test errors to stderr so users don't have to `cat` or open test failure log files when test
 # fail. This makes the log noiser in exchange for reducing the time-to-feedback on test failures for
diff --git a/lib/tests/bazelrc_presets/all/correctness.bazelrc b/lib/tests/bazelrc_presets/all/correctness.bazelrc
index d9ef2e5bc..56fad5fd5 100644
--- a/lib/tests/bazelrc_presets/all/correctness.bazelrc
+++ b/lib/tests/bazelrc_presets/all/correctness.bazelrc
@@ -11,7 +11,6 @@ build --noremote_upload_local_results
 # Developers should tag targets with `tags=["requires-network"]` to opt-out of the enforcement.
 # Docs: https://bazel.build/reference/command-line-reference#flag--sandbox_default_allow_network
 build --sandbox_default_allow_network=false
-test --sandbox_default_allow_network=false
 
 # Warn if a test's timeout is significantly longer than the test's actual execution time.
 # Bazel's default for test_timeout is medium (5 min), but most tests should instead be short (1 min).
diff --git a/lib/tests/bazelrc_presets/all/performance.bazelrc b/lib/tests/bazelrc_presets/all/performance.bazelrc
index bc11a3f43..8705c2a6e 100644
--- a/lib/tests/bazelrc_presets/all/performance.bazelrc
+++ b/lib/tests/bazelrc_presets/all/performance.bazelrc
@@ -41,8 +41,6 @@ build --experimental_reuse_sandbox_directories
 # author.
 # Docs: https://bazel.build/reference/command-line-reference#flag--legacy_external_runfiles
 build --nolegacy_external_runfiles
-run --nolegacy_external_runfiles
-test --nolegacy_external_runfiles
 
 # Some actions are always IO-intensive but require little compute. It's wasteful to put the output
 # in the remote cache, it just saturates the network and fills the cache storage causing earlier
@@ -52,3 +50,8 @@ test --nolegacy_external_runfiles
 # In some cases you may need to patch rulesets to add a mnemonic to actions that don't have one.
 # https://bazel.build/reference/command-line-reference#flag--modify_execution_info
 build --modify_execution_info=PackageTar=+no-remote
+
+# Build runfiles symlink forests for binaries and tests on demand rather than at build time.
+# This reduces the overhead incurred when running a binary or test isn't required.
+# Docs: https://bazel.build/docs/user-manual#build-runfile-links
+build --nobuild_runfile_links