URGENT: Issue with merging and integration of 6 seurat objects

[PantehaK]

Brief Overview
I'm working with six Seurat objects corresponding to 6 reactions/batches (GEM-A, GEM-B, GEM-1, GEM-2, GEM-3, GEM-4) each of these contain x4 samples multiplexed together (so a total of 24 samples). I want to merge and integrate by sample rather than batch. However, the objects contain batch-separated layers (e.g., 'counts.GEM-A', 'data.GEM-A', etc.) within the 'RNA' assay. After merging, I intend to run NormalizeData(), ScaleData(), and RunPCA() before proceeding with integration as per this vignette https://satijalab.org/seurat/articles/seurat5_integration.

Problems Encountered
Failure to Rebuild Assays using JoinLayers() When I try to combine all split layers within the 'RNA' assay using JoinLayers(), I encounter the following error:
Error in fn():
! Cannot add new cells with [[<-
Error during wrapup: long vectors not supported yet: ../../src/include/Rinlinedfuns.h:537

Even when requesting up to 500GB of memory, this approach fails. This error occurs particularly when I try to JoinLayers() on all six objects together. However, the same approach works perfectly fine when I merged four objects together (total 16 samples), suggesting the error is related to memory exhaustion or the number of split layers.

Splitting Merged Object with SplitObject() Fails I attempted to directly split the merged object by sample using:
split_object <- SplitObject(merged, split.by = 'sample')
This results in:
RNA assay doesn't leave any cells, so it is removed
Error in subset.Seurat(x = object, cells = cells) :
No cells left in the default assay, please change the default assay

Errors Occur When Attempting Normalisation and other processing
Even if I proceed WITHOUT splitlayers and keep the merged object as is, NormalizeData(), FindVariableFeatures(), ScaleData(), etc., all return similar errors:
Error in fn():
! Cannot add new cells with [[<-

Attempting to process only the counts layer via:
merged <- NormalizeData(merged, normalization.method = 'LogNormalize', layers = 'counts')
Fails with the same error.

What I Have Tried

• Rebuilding the RNA assay for each object independently with JoinLayers() before merging.
• Removing scale.data layers before merging to reduce memory usage.
• Processing each object separately before merging.
• Specifying layers = 'counts' during normalization.
• Attempting to split the merged object directly using SplitObject().

Request
I am hoping you could help me with how to properly handle my merged objects, normalising, and splitting the merged object by sample. I would greatly appreciate any insights or suggestions you can provide. Do I need to break my data down into smaller pieces? How should I go about combining them together for analysis then?

My main goal is to be able to integrate all 6 runs (24 samples) together using Harmony and carry on with DEX