[Improvement]: Improve worker message passing to reduce memory leak

[HindujaB]

Description

$subject

Describe your problem(s)

The following code which passes 10KB json data between 3 workers increases the memory usage over time.
The heap dump shows a large count of WorkerDataChannel objects.

import ballerina/time;

// json payload = some 10KB json value
public function main() returns error? {
    time:Utc utc = time:utcAddSeconds(time:utcNow(), 1800); // 30 mins running
    while (utc >= time:utcNow()) {
        check testWorkerMessaging();
    }
}

isolated function testWorkerMessaging() returns error? {

    worker w1 returns error? {
        json res1 = clientFunc1(payload);
        res1 -> w2;
    }

    worker w2 returns error? {
        json res1 = check <- w1;
        json res2 = clientFunc2(res1);
        res2 -> w3;
    }

    worker w3 returns error? {
        json res2 = check <- w2;
        json res3 = clientFunc3(res2);
        res3 -> function;
    }

    json res3 = check <- w3;
    check wait w3;
    return ();
}

type Rec record {|
    int i;

|};

isolated function clientFunc1(json payload1) returns json {
    return payload1;
}

isolated function clientFunc2(json payload1) returns json {
    return payload1;
}

isolated function clientFunc3(json payload1) returns json {
    return payload1;
}

Memory graph :

Heap dump :

Heap dump file :
https://drive.google.com/file/d/1dXK7C2tZiPxXkHsfhEYhXgRaG7y_Datw/view?usp=sharing

Describe your solution(s)

Clean up the channels & channel details.

Related area

-> Runtime

Related issue(s) (optional)

No response

Suggested label(s) (optional)

No response

Suggested assignee(s) (optional)

No response

[HindujaB]

After removing the redundant WorkerDataChannel and ChannelDetails instances, the memory usage is reduced to the following.
Memory graph :

Heap dump :

[HindujaB]

This is now available in 2201.9.0.