[CI] PushQueriesIT testCaseInsensitiveEquality {auto} failing

[elasticsearchmachine]

Build Scans:

• elasticsearch-periodic #8441 / graalvm-ce17_checkpart3_java-matrix
• elasticsearch-periodic-platform-support #8398 / windows-2019_checkpart3_platform-support-windows

Reproduction Line:

./gradlew ":x-pack:plugin:esql:qa:server:single-node:javaRestTest" --tests "org.elasticsearch.xpack.esql.qa.single_node.PushQueriesIT.testCaseInsensitiveEquality {auto}" -Dtests.seed=77829EE3BC96499B -Dtests.locale=sv-AX -Dtests.timezone=America/Argentina/Catamarca -Druntime.java=17

Applicable branches:
8.19

Reproduces locally?:
N/A

Failure History:
See dashboard

Failure Message:

java.lang.AssertionError: Expected a map containing
      took: <16>
is_partial: <false>
   profile: a map containing
     drivers: <[{task_description=data, sleeps={last=[], counts={}, first=[]}, cpu_nanos=1685991, operators=[{operator=LuceneSourceOperator[maxPageSize = 4854, remainingDocs = 2147483646], status={processed_slices=1, processed_shards=[test:0], total_slices=1, slice_max=0, current=2147483647, slice_min=0, rows_emitted=1, pages_emitted=1, partitioning_strategies={test:0=SHARD}, processing_nanos=157572, processed_queries=[*:*], slice_index=0}}, {operator=ValuesSourceReaderOperator[fields = [test]], status={pages_processed=1, rows_emitted=1, process_nanos=1228317, rows_received=1, readers_built={test:column_at_a_time:null=1, stored_fields[requires_source:true, fields:0, sequential: false]=1, test:row_stride:BlockSourceReader.Bytes=1}}}, {operator=FilterOperator[evaluator=InsensitiveEqualsConstantEvaluator[lhs=Attribute[channel=1], rhs=initial state: 0
state 0 [reject]:
]], 
[truncated]

Issue Reasons:

• [8.19] 2 failures in test testCaseInsensitiveEquality {auto} (0.5% fail rate in 387 executions)

Note:
This issue was created using new test triage automation. Please report issues or feedback to es-delivery.

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[nik9000]

We've had some issues with ML that blow up in this test because of it's use of semantic_text. I'm investigating to see if this is that.

[nik9000]

Second failure is a socket timeout indexing text fields. That doesn't make a lot of sense and I'm asking around. I don't have a build artifact to look at.

Other failure was:

    values: expected <[[]]> but was <[]>

Checking that one.

Neither look like the ML failures I've sent to our ml friends.

[nik9000]

Other failure was:

    values: expected <[[]]> but was <[]>

Checking that one.

That's a lack of backport of #127532 . Doing that now.

[nik9000]

The second failure tracks down to:

cluster state update executor did not preserve version numbers: [update model deployment]

Which is the ML failure that's reported elsewhere.

[nik9000]

Backports that are needed:
#128578
#128580

I'm going to mute the semantic_text field tests and attach that to #128511