Spark streaming (merge into) iceberg table concurrent write with compaction job

[2MD]

Query engine

Iceberg version 1.7.1
Spark version 3.3.2

Question

We have spark streaming application A1 which in every microbatch does:

s"""
       |MERGE INTO table as t
       |USING (select * from $tempViewName) as s
       |ON $joinCondition
       |WHEN MATCHED AND s.$versionColumnName > t.$versionColumnName THEN UPDATE SET *
       |WHEN NOT MATCHED THEN INSERT *
       |"""

We can update any data file.
Table "table" without partition.

And we have some spark batch application A2 which does:
remove old snapshot, rewrite manifest , compaction (binpack and sometimes z-order), rewrite position delete files, delete orphan files.
For this table.

How we can avoid concurrent troubles between two applications?

(We are still thinking about launching A2 inside mircobatch A1... but is not the best solution)