diff --git a/processes/build_references/build_reference.nf b/processes/build_references/build_reference.nf
index 451d0dbe..1ad752c7 100755
--- a/processes/build_references/build_reference.nf
+++ b/processes/build_references/build_reference.nf
@@ -14,7 +14,7 @@ readlengths = params.readlength.tokenize(',')
 
 process bwa {
 
-  publishDir "${params.outdir}/bwa"
+  publishDir "${params.outdir}/bwa", mode: "link"
 
   input:
   file genome from file(params.genome)
@@ -29,7 +29,7 @@ process bwa {
 }
 
 process samtools {
-  publishDir "${params.outdir}/samtools"
+  publishDir "${params.outdir}/samtools", mode: "link"
 
   input:
   file genome from file(params.genome)
@@ -61,7 +61,7 @@ process bowtie_index {
 
 process mappability {
 
-  publishDir "${params.outdir}/annotations"
+  publishDir "${params.outdir}/annotations", mode: "link"
 
   input:
   file genome from file(params.genome)
@@ -85,7 +85,7 @@ process mappability {
 
 process density {
 
-  publishDir "${params.outdir}/densities"
+  publishDir "${params.outdir}/densities", mode: "link"
 
   input:
   file fai
@@ -104,7 +104,7 @@ process density {
 
 process chrom_sizes {
 
-  publishDir "${params.outdir}/hotspot2"
+  publishDir "${params.outdir}/hotspot2", mode: "link"
 
   input:
   file fai
@@ -120,7 +120,7 @@ process chrom_sizes {
 
 process chrom_info {
 
-  publishDir "${params.outdir}/annotations"
+  publishDir "${params.outdir}/annotations", mode: "link"
 
   input:
   file chrom_sizes
@@ -138,7 +138,7 @@ process hotspot2 {
 
   container "fwip/hotspot2:latest"
 
-  publishDir "${params.outdir}/hotspot2"
+  publishDir "${params.outdir}/hotspot2", mode: "link"
 
   input:
   file chrom_sizes
@@ -161,7 +161,7 @@ process hotspot2 {
 }
 
 process nuclear_center_sites {
-  publishDir "${params.outdir}/hotspot2"
+  publishDir "${params.outdir}/hotspot2", mode: "link"
 
   input:
   file center_sites
diff --git a/processes/bwa/aggregate/basic.nf b/processes/bwa/aggregate/basic.nf
index c25443d7..bd36dfd1 100644
--- a/processes/bwa/aggregate/basic.nf
+++ b/processes/bwa/aggregate/basic.nf
@@ -48,7 +48,7 @@ process merge {
   output:
   file 'merged.bam' into merged
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   script:
   """
@@ -59,7 +59,7 @@ process merge {
 // TODO: single end
 process dups {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   label 'high_mem'
 
   input:
@@ -97,7 +97,7 @@ marked_bam.into { bam_for_counts; bam_for_adapter_counts; bam_for_filter; bam_fo
 process filter {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file bam from bam_for_filter
@@ -135,7 +135,7 @@ process filter_nuclear {
 
 process macs2 {
   label "macs2"
-  publishDir "${params.outdir}/peaks_macs2"
+  publishDir "${params.outdir}/peaks_macs2", mode: "link"
   scratch false
 
   when:
@@ -161,7 +161,7 @@ process macs2 {
 process hotspot2 {
   label "modules"
 
-  publishDir "${params.outdir}"
+  publishDir "${params.outdir}", mode: "link"
   container "fwip/hotspot2:latest"
 
   when:
@@ -215,7 +215,7 @@ process hotspot2 {
 
 process spot_score {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file(bam) from bam_for_spot_score
@@ -256,7 +256,7 @@ process spot_score {
 
 process bam_counts {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file(bam) from bam_for_counts
@@ -274,7 +274,7 @@ process bam_counts {
 
 process count_adapters {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file(bam) from bam_for_adapter_counts
@@ -292,7 +292,7 @@ process count_adapters {
 
 process preseq {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   input:
   file nuclear_bam
 
@@ -318,7 +318,7 @@ process preseq {
 process cutcounts {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file(fai) from file("${params.genome}.fai")
@@ -366,7 +366,7 @@ process cutcounts {
 process density {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   label 'high_mem'
 
   input:
@@ -419,7 +419,7 @@ process density {
 
 process multimapping_density {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   label 'modules'
   label 'high_mem'
 
@@ -498,7 +498,7 @@ process multimapping_density {
 
 process normalize_density {
   label "modules"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set(file(filtered_bam), file(density)) from to_normalize
@@ -541,7 +541,7 @@ process normalize_density {
 process insert_sizes {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file nuclear_bam from bam_for_inserts
@@ -571,7 +571,7 @@ process insert_sizes {
 process motif_matrix {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file hotspot_calls
@@ -595,7 +595,7 @@ process motif_matrix {
 process closest_features {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file hotspot_calls
@@ -639,7 +639,7 @@ process closest_features {
 process differential_hotspots {
   label "modules"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file bam from bam_for_diff_peaks
@@ -683,7 +683,7 @@ process differential_hotspots {
 process learn_dispersion {
 
   label "footprints"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   memory = '8 GB'
   cpus = 8
@@ -805,7 +805,7 @@ process working_tracks {
   memory = '32 GB'
   cpus = 1
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file merged_interval
@@ -833,7 +833,7 @@ process compute_footprints {
   memory = '8 GB'
   cpus = 1
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set file(merged_interval), val(threshold) from merged_interval.combine(thresholds)
@@ -861,7 +861,7 @@ process compute_footprints {
 process plot_footprints {
 
   label "footprints"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file model from to_plot
diff --git a/processes/bwa/process_bwa_paired_trimmed.nf b/processes/bwa/process_bwa_paired_trimmed.nf
index e3f48ce5..88c8f873 100755
--- a/processes/bwa/process_bwa_paired_trimmed.nf
+++ b/processes/bwa/process_bwa_paired_trimmed.nf
@@ -308,7 +308,7 @@ process mark_duplicates {
 
   label "high_mem"
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file(merged_bam) from merged_bam
@@ -352,7 +352,7 @@ if (params.UMI)
 
 process filter_bam_to_unique {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file marked_bam
@@ -396,7 +396,7 @@ process bam_counts {
  */
 process insert_size {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set file(bam), file(bai) from bam_for_insert
@@ -429,7 +429,7 @@ process insert_size {
 
 process spot_score {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set file(bam), file(bai) from bam_for_spot
@@ -484,7 +484,7 @@ win = 75
 bini = 20
 process density_files {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set file(bam), file(bai) from bam_for_density
@@ -530,7 +530,7 @@ process density_files {
  */
 process total_counts {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file 'fastqcounts*' from fastq_counts.collect()
diff --git a/processes/rna-star/aggregation/cufflinks_featurecounts.nf b/processes/rna-star/aggregation/cufflinks_featurecounts.nf
index e44c8bd2..92fe071e 100644
--- a/processes/rna-star/aggregation/cufflinks_featurecounts.nf
+++ b/processes/rna-star/aggregation/cufflinks_featurecounts.nf
@@ -69,7 +69,7 @@ workflow RNA_AGG {
 
 process merge_transcriptome_bam {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "samtools/1.12"
   input:
     // Assume sorted by coord
@@ -90,7 +90,7 @@ process merge_transcriptome_bam {
 }
 
 process merge_genome_bam {
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "samtools/1.12"
   input:
     // Assume sorted by coord
@@ -114,7 +114,7 @@ process merge_genome_bam {
 
 
 process remove_duplicate_reads {
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "jdk/2.8.1", "picard/2.8.1", "samtools/1.12"
   label 'high_mem'
   input:
@@ -152,7 +152,7 @@ process remove_duplicate_reads {
 }
 
 process mark_duplicate_reads {
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   module "jdk/2.8.1", "picard/2.8.1", "samtools/1.12"
   input:
@@ -177,7 +177,7 @@ process mark_duplicate_reads {
 
 
 process bam_to_fastq {
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "samtools/1.12"
   
   input:
@@ -202,7 +202,7 @@ process bam_to_fastq {
 process density {
 
   module "STAR/2.4.2a", "bedops/2.4.19", "htslib/1.6.0"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   input:
     path(input_bam)
     path("ref/*")
@@ -260,7 +260,7 @@ process density {
 
 process cufflinks {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "cufflinks/2.2.1", "R/3.2.5", "anaquin/2.0.1"
 
   input:
@@ -300,7 +300,7 @@ process cufflinks {
 
 process stringtie {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "stringtie/1.3.4d"
   input:
     path(input_bam)
@@ -321,7 +321,7 @@ process stringtie {
 
 process feature_counts {
 
-    publishDir params.outdir
+    publishDir params.outdir, mode: "link"
     module "subread/1.5.1"
 
     input:
@@ -343,7 +343,7 @@ process feature_counts {
 
 process kallisto {
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "kallisto/0.43.1", "anaquin/2.0.1"
 
   input:
@@ -369,7 +369,7 @@ process kallisto {
 
 process kallisto_advanced {
   
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   module "kallisto/0.43.1", "anaquin/2.0.1"
 
   input:
@@ -397,7 +397,7 @@ process kallisto_advanced {
 process anaquin {
 
   module "samtools/1.12", "anaquin/2.0.1", "kallisto/0.43.1", "R/3.2.5"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
     path input_bam
@@ -457,7 +457,7 @@ process anaquin {
 process insert_sizes {
 
   module "jdk", "picard/2.9.0", "R/3.2.5"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
     path input_bam
@@ -474,7 +474,7 @@ process insert_sizes {
 process rna_metrics {
 
   module "jdk", "picard/2.9.0"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
     path input_bam
@@ -495,7 +495,7 @@ process rna_metrics {
 process adapter_count {
 
   module "bwa", "samtools", "jdk", "picard/2.9.0"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
   input:
     path input_bam
 
@@ -513,7 +513,7 @@ process adapter_count {
 process ribosomal_count {
 
   module "bowtie/1.0.0"
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
     tuple path(r1_fq), path(r2_fq)
diff --git a/processes/rna-star/modules/star.nf b/processes/rna-star/modules/star.nf
index a88ba66c..81664dbb 100644
--- a/processes/rna-star/modules/star.nf
+++ b/processes/rna-star/modules/star.nf
@@ -8,7 +8,7 @@ process star {
   cpus params.star_threads
   module 'STAR/2.4.2a', 'samtools/1.7'
 
-  publishDir params.outdir, enabled: params.publish
+  publishDir params.outdir, enabled: params.publish, mode: "link"
 
   label 'high_mem'
 
diff --git a/processes/rna-star/rna_star.nf b/processes/rna-star/rna_star.nf
index a12dd234..cab6e12d 100755
--- a/processes/rna-star/rna_star.nf
+++ b/processes/rna-star/rna_star.nf
@@ -49,7 +49,7 @@ process star {
   cpus params.star_threads
   module 'STAR', 'samtools/1.7', 'gcc/4.7.2'
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   set file(r1), file(r2) from trimmed
@@ -102,7 +102,7 @@ process star_bedgraph {
 
   module 'STAR'
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file coordBam from coordinateBam
@@ -128,7 +128,7 @@ process rsem {
 
   module 'perl/5.16.3', 'RSEM/1.2.30'
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file transcriptBam
@@ -158,7 +158,7 @@ process rsem_plot {
 
   module 'perl/5.16.3', 'RSEM/1.2.30', 'R/3.2.5'
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   input:
   file "Quant.stat/*" from rsem_quant
diff --git a/processes/rna-star/star_only.nf b/processes/rna-star/star_only.nf
index 9c6b7cbc..bd098ebf 100755
--- a/processes/rna-star/star_only.nf
+++ b/processes/rna-star/star_only.nf
@@ -49,7 +49,7 @@ process star {
   cpus params.star_threads
   module 'STAR', 'samtools/1.7'
 
-  publishDir params.outdir
+  publishDir params.outdir, mode: "link"
 
   label 'high_mem'
 
diff --git a/scripts/utility/estimate_nf_clean.sh b/scripts/utility/estimate_nf_clean.sh
new file mode 100755
index 00000000..3381e4a6
--- /dev/null
+++ b/scripts/utility/estimate_nf_clean.sh
@@ -0,0 +1,12 @@
+#!/bin/bash
+
+for dir in "$@" ; do
+  [[ -d "$dir/work" ]] || continue
+  dir=$(readlink -f "$dir")
+  find "$dir/work" -type f \
+    | grep -v -F -f <( \
+      find "$dir"/output* -type l -print0 \
+      | xargs --no-run-if-empty -0 readlink -f \
+    ) \
+    | xargs --no-run-if-empty du -scb | tail -n1  | sed "s!total!$dir!"
+done
diff --git a/scripts/utility/nextflow_clean.sh b/scripts/utility/nextflow_clean.sh
new file mode 100755
index 00000000..d648e297
--- /dev/null
+++ b/scripts/utility/nextflow_clean.sh
@@ -0,0 +1,221 @@
+#!/bin/bash
+
+
+# 1   visit directory containing work/ dir
+# 2   find all symlinks within top-level dir (and subdirs), excluding those in work dir.
+# 3   for each symlink found:
+# 3A    get the target of the symlink, resolving any symlink chains (e.g: the symlink may point to another symlink, etc)
+# 3B    if the symlink does not point into a work/ directory, skip it.
+# 3C    make a copy of the link for backup purposes in case we get interrupted (e.g: mysymlink.bak)
+# 3D    replace the symlink with a hard link to the real target
+# 3E    remove the symlink backup
+# 4   repeat step 2, exiting with error if any symlinks remain.
+# 5   remove work directory
+# 6   set LIMS status
+
+
+set -euo pipefail
+
+AGG_DIR_REGEX=/net/seq/data/aggregations/LN[0-9]+/aggregation-[0-9]+
+
+#########################
+
+function usage(){
+  echo "Usage: $0 [-f] dir [dirs...]"  >&2
+  exit 2
+}
+
+# Parse arguments
+FORCE=
+while getopts ":fh" arg; do
+  case $arg in
+    h)
+      usage
+      ;;
+    f)
+      FORCE=TRUE
+      ;;
+    *)
+      usage
+      ;;
+
+  esac
+done
+shift $((OPTIND -1))
+[[ -z "$*" ]] && usage
+
+# Print arguments to stderr (printf style format string) and exit
+DIE_F(){
+  printf "ERROR: "
+  # shellcheck disable=SC2059
+  printf "$@"
+  printf "\n"
+  exit 1
+}
+INFO_F(){
+  printf "INFO: "
+  # shellcheck disable=SC2059
+  printf "$@"
+  printf "\n"
+}
+
+RUN_IF_FORCE(){
+  if [[ "$FORCE" == TRUE ]] ; then
+    printf "INFO: Running:' "
+    printf "%q " "$@"
+    printf "'\n"
+    "$@"
+  else
+    printf "INFO: Dry-run, would have run: '"
+    printf "%q " "$@"
+    printf "'\n"
+  fi
+}
+
+lims_patch_by_url() {
+  /usr/bin/curl \
+  --request PATCH \
+  --data "$2" \
+  "$1" \
+  -H "Authorization: Token $LIMS_API_TOKEN" \
+  -k \
+  2>/dev/null
+}
+
+# $1: url, $2: field=val
+lims_patch () {
+  lims_patch_by_url "$LIMS_API_URL/$1" "$2"
+}
+
+lock_aggregation() {
+  local aggdir
+  local aggid
+  aggdir=$(readlink -f $1)
+  if  [[ ! "$aggdir" =~ $AGG_DIR_REGEX ]] ; then
+    INFO_F "Not an aggregation directory, not locking"
+    return
+  fi
+
+  # Get aggregation ID from directory name
+  aggid=${aggdir/*aggregation-/}
+  if [[ ! $aggid  =~ ^[0-9]+$ ]] ; then
+    DIE_F "Couldn't parse aggregation directory '%s' ('%s') into aggid (got '%s')" "$1" "$aggdir" "$aggid"
+  fi
+
+  lims_patch "aggregation/$aggid/" 'locked=True'
+}
+
+# Find output symlinks in a directory
+# Expects them to be in subdirectories starting with 'output'
+find_output_symlinks(){
+  # find all symlinks within top-level dir (and subdirs), excluding those in work dir.
+  find "$1" -maxdepth 1 -name 'output*' -type d -print0 \
+    | xargs --no-run-if-empty -0 -I '{}' find '{}' -type l
+}
+
+has_broken_symlink(){
+  local broken_symlink
+  broken_symlink=$(find "$1" -maxdepth 1 -name 'output*' -type d -print0 \
+    | xargs --no-run-if-empty -0 -I '{}' find '{}' -type l -xtype l -print -quit)
+  [[ -n "$broken_symlink" ]]
+}
+
+# Following symlinks, see if the two arguments point to the same region of the disk.
+is_same_inode(){
+  [[ "$1" -ef "$2" ]]
+}
+
+process_dir() {
+  local dir=$1
+  # Start subshell to isolate our changes/work
+  # Also redirecting logs
+  (
+    INFO_F "Processing dir '%s'" "$dir"
+    # Make sure dir exists
+    [[ -d "$dir" ]] || DIE_F "Directory does not exist: '%s'" "$dir"
+    # It should also contain a workdir
+    [[ -d "$dir/work" ]] || DIE_F "Work directory does not exist: '%s/work'" "$dir"
+
+    set -euo pipefail  # This is inherited, but in case somebody changes the top declaration.
+
+    # 1) visit directory containing work/ dir
+    cd "$dir"
+
+    # 1.5) Check for broken symlinks
+    has_broken_symlink . &&
+      DIE_F "Broken symlink detected. dir=%s" "$dir"
+
+    # 2) find symlinks
+    local symlinks=()
+    readarray -t symlinks < <( find_output_symlinks . )
+
+    # Check 
+    # https://stackoverflow.com/questions/7577052/bash-empty-array-expansion-with-set-u
+    # shellcheck disable=SC2199
+    if [[ -z ${symlinks[@]+"${symlinks[@]}"} ]] ; then
+      INFO_F "No input symlinks found. dir=%s" "$dir"
+      symlinks=()
+    else
+
+      # 3)
+      for symlink in "${symlinks[@]}" ; do
+        # 3A) Get symlink target
+        local target
+        target=$(readlink -f "$symlink")
+        # 3B) Die if not in work directory
+        [[ "$target" =~ /work/ ]] ||
+          DIE_F "Target not in workdir. dir='%s';symlink='%s';target='%s'" "$dir" "$symlink" "$target"
+
+        # 3?) Make sure symlink isn't broken
+        [[ -e "$target" ]] ||
+          DIE_F "Target not found. dir='%s';symlink='%s';target='%s'" "$dir" "$symlink" "$target"
+
+        is_same_inode "$target" "$symlink" ||
+          DIE_F "Somehow, target & symlink inodes differ. dir='%s';symlink='%s';target='%s'" "$dir" "$symlink" "$target"
+
+        # 3C) make a copy of the link for backup purposes in case we get interrupted (e.g: mysymlink.bak)
+        local symlink_bak=$symlink.bak
+        RUN_IF_FORCE cp --no-dereference "$symlink" "$symlink_bak"
+
+        # 3D) Create hard link to real target, replacing soft link
+        local hardlink=$symlink
+        RUN_IF_FORCE ln -f --logical "$target" "$hardlink"
+
+        # Triple-check, make sure the hard link and backup symlink point to the same place
+        if [[ "$FORCE" == TRUE ]] ; then
+          is_same_inode "$symlink_bak" "$hardlink" ||
+            DIE_F "Somehow, the symlink and hard link point to different files. hardlink='%s', target='%s'" "$hardlink" "$target"
+        fi
+
+        # 3E) remove backup symlink
+        RUN_IF_FORCE rm "$symlink_bak"
+      done
+    fi
+
+    # Only run check for symlinks if we deleted
+    if [[ $FORCE == TRUE ]] ; then
+      # 4)
+      local remaining_symlinks
+      remaining_symlinks=$(find_output_symlinks .)
+      [[ -z "$remaining_symlinks" ]] ||
+        DIE_F "Some symlinks remain, not removing workdir. dir='%s';remaining='%q'" "$dir" "${remaining_symlinks[*]}"
+    else
+      INFO_F "Dry-run, skipping the check for remaining symlinks"
+    fi
+
+    RUN_IF_FORCE rm -r --one-file-system --preserve-root "work"
+
+    RUN_IF_FORCE lock_aggregation .
+
+    INFO_F "Done. dir='%s'" "$dir"
+
+  ) | tee "$dir"/freezing.log
+}
+
+for directory in "$@" ; do
+  if [[ -d "$directory/work" ]]  ; then
+    process_dir "$(readlink -f $directory)"
+  else
+    INFO_F "Dir '%s' does not contain a workdir" "$directory"
+  fi
+done