diff --git a/website/user_faq.md b/website/user_faq.md
index cfbf942..fdbfb17 100644
--- a/website/user_faq.md
+++ b/website/user_faq.md
@@ -114,6 +114,12 @@ path or set an environment variable accordingly.
 
 [⤴ _**back to Content**_](#content)
 
+## Julia unexpectedly killed for exceeding the requested memory limit
+
+If a job has non-exclusive access to a node and has a memory limit that is lower than the total memory of the node, set the `--heap-size-hint` command line option to an appropriate value when starting Julia in the job script, e.g. `julia --heap-size-hint=4G my_script.jl` if you have requested a memory limit of 4G for running `my_script.jl`. This communicates the memory limit to Julia's garbage collector to enable more aggressive garbage collection when the memory limit is approached.
+
+[⤴ _**back to Content**_](#content)
+
 ## I get memory-related issues when using CUDA.jl on a HPC cluster
 
 Try setting `JULIA_CUDA_MEMORY_POOL=none` (see the [CUDA.jl documentation](https://cuda.juliagpu.org/stable/usage/memory/#Memory-pool) for more information).
diff --git a/website/user_gettingstarted.md b/website/user_gettingstarted.md
index eb9c3d9..b5b6214 100644
--- a/website/user_gettingstarted.md
+++ b/website/user_gettingstarted.md
@@ -39,7 +39,7 @@ You want to choose a file system with the following properties
 * good (parallel) I/O
 * no automatic deletion of unused files (or otherwise you have to find a workaround)
 
-**On most clusters these criterion are best fit on a parallel file system (often `$SCRATCH`).** In this case, you should put `JULIA_DEPOT_PATH=$SCRATCH/.julia` into your `.bashrc`.
+**On most clusters these criterion are best fit on a parallel file system (often `$SCRATCH`).** In this case, you should put `JULIA_DEPOT_PATH=$SCRATCH/.julia` into your `.bashrc` (and your job scripts if `.bashrc` is not loaded by non-interactive jobs).
 
 **Note:** If the last point (automatic deletion of unused files) is an issue for you, a pragmatic workaround could be a cronjob that touches all files in the Julia depot every once in a while.
 
@@ -56,4 +56,4 @@ To make Julia produce efficient code that works on different CPUs, you need to s
 
 You can get the CPU target name for the current system with `Sys.CPU_NAME`. For more information, see [this section of the Julia documentation](https://docs.julialang.org/en/v1/manual/environment-variables/#JULIA_CPU_TARGET) and [this section of the developer documentation](https://docs.julialang.org/en/v1/devdocs/sysimg/#Specifying-multiple-system-image-targets).
 
-[⤴ _**back to Content**_](#content)
\ No newline at end of file
+[⤴ _**back to Content**_](#content)