diff --git a/pipelines/convergenceratio/convergenceratio.nf b/pipelines/convergenceratio/convergenceratio.nf
index 0cd55fd7..bcedcc5c 100644
--- a/pipelines/convergenceratio/convergenceratio.nf
+++ b/pipelines/convergenceratio/convergenceratio.nf
@@ -1,5 +1,5 @@
 
-include { all_by_all_blast; blastreduce; blastreduce_transcode_fasta; condense_redundant; create_blast_db; restore_condensed; split_fasta } from "../shared/nextflow/blast.nf"
+include { all_by_all_blast; blastreduce; blastreduce_transcode_fasta; condense_redundant; create_blast_db; restore_condensed; sort_and_split_fasta } from "../shared/nextflow/blast.nf"
 include { unzip_ssn } from "../shared/nextflow/util.nf"
 include { compute_clusters; get_conv_ratio_table; get_id_list; get_ssn_id_info } from "../shared/nextflow/color_workflow.nf"
 
@@ -131,7 +131,7 @@ workflow {
 
     fasta_lengths_parquet = blastreduce_transcode_fasta(cluster_fasta)
 
-    fasta_shards = split_fasta(reduced_fasta.fasta_file)
+    fasta_shards = sort_and_split_fasta(reduced_fasta.fasta_file)
 
     blast_input = blast_databases.combine(fasta_shards.transpose(), by: 0)
 
diff --git a/pipelines/est/subworkflows/all_by_all.nf b/pipelines/est/subworkflows/all_by_all.nf
index 78b8e4a3..3c5812cb 100644
--- a/pipelines/est/subworkflows/all_by_all.nf
+++ b/pipelines/est/subworkflows/all_by_all.nf
@@ -1,11 +1,14 @@
 
-include { all_by_all_blast; blastreduce; blastreduce_transcode_fasta; condense_redundant; create_blast_db; restore_condensed; split_fasta } from "../../shared/nextflow/blast.nf"
+include { all_by_all_blast; blastreduce; blastreduce_transcode_fasta; condense_redundant; create_blast_db; restore_condensed; sort_and_split_fasta } from "../../shared/nextflow/blast.nf"
 
 workflow ALL_BY_ALL {
     take:
         original_fasta
 
     main:
+        // For stats computation later; using the original fasta file
+        fasta_lengths_parquet = blastreduce_transcode_fasta(original_fasta)
+
         // Cluster redundant sequences for BLAST computation (formerly known as multiplex).
         // Only performed when input sequences are from Uniprot, since sequence sets from
         // UniRef90 and UniRef50 are already sequence-unique.
@@ -21,11 +24,8 @@ workflow ALL_BY_ALL {
         // Create BLAST database
         blastdb = create_blast_db(blast_input_fasta)
 
-        // For stats computation later
-        fasta_lengths_parquet = blastreduce_transcode_fasta(original_fasta)
-
         // All-by-all BLAST
-        fasta_shards = split_fasta(blast_input_fasta)
+        fasta_shards = sort_and_split_fasta(blast_input_fasta)
 
         blast_input = blastdb.combine(fasta_shards.transpose(), by: 0)
         blast_fractions = all_by_all_blast( blast_input ).groupTuple()
diff --git a/pipelines/shared/nextflow/blast.nf b/pipelines/shared/nextflow/blast.nf
index 2162bbbf..305baf8e 100644
--- a/pipelines/shared/nextflow/blast.nf
+++ b/pipelines/shared/nextflow/blast.nf
@@ -182,7 +182,7 @@ process restore_condensed {
 }
 
 
-process split_fasta {
+process sort_and_split_fasta {
     input:
         tuple val(fid), path(fasta_file)
 
@@ -192,10 +192,16 @@ process split_fasta {
     script:
     """
     mkdir parts
-    seqkit split2 \
+    seqkit sort --by-length \
+        --reverse \
+        --two-pass \
+        --threads ${task.cpus} \
         ${fasta_file} \
+        | seqkit split2 \
+        --threads ${task.cpus} \
         -p ${params.num_fasta_shards} \
-        --out-dir parts
+        --out-dir parts \
+        --out-prefix "all_sequences.fasta_"
     """
 }