Hadoop Streaming API's compatibility with Apache Pig is crucial in optimizing data processing, especially for handling large datasets. Pig allows developers to express data transformations using a high-level scripting language, making it easier to work with complex data processing tasks.

Apache Pig's Nested Data feature enables the processing of nested data structures, providing flexibility in handling complex data types within the Hadoop ecosystem. It allows users to work with data that has varying and nested structures without predefined schemas.

The primary purpose of Apache Pig in the Hadoop ecosystem is data analysis. It provides a platform for creating and executing data analysis programs using a high-level scripting language called Pig Latin, making it easier to work with large datasets.

To enhance performance, Apache Spark is often configured in Hadoop clusters to manage large-scale data processing. Spark provides in-memory processing capabilities and a high-level API, making it suitable for iterative algorithms and interactive data analysis.

The crucial configuration for specifying the output key and value types in a MapReduce job is map.output.value.class. This configuration defines the data types emitted by the Mapper.

When dealing with intermittent failures, log analysis is the most effective monitoring approach for diagnosis. Examining Hadoop logs can provide insights into error messages, stack traces, and events that occurred during job execution, helping troubleshoot and identify the root cause of failures.

The replication factor in Hadoop impacts cluster capacity planning by improving fault tolerance. Higher replication ensures data availability even if some nodes fail. However, it comes at the cost of increased storage requirements. Capacity planning needs to balance fault tolerance with storage efficiency.

The choice of file format in Apache Hive plays a crucial role in optimizing query performance. ORC (Optimized Row Columnar) is specifically designed for high-performance analytics by organizing data in a way that minimizes I/O and improves compression, leading to faster query execution.

Sqoop is the tool in Hadoop specialized for bulk data transfer between Hadoop and relational databases. It simplifies the process of importing and exporting data, allowing seamless integration of data stored in databases with the Hadoop ecosystem.

Replication in Hadoop development is crucial for ensuring data integrity and fault tolerance. It involves creating duplicate copies of data blocks and storing them across different nodes in the cluster, reducing the risk of data loss and improving fault tolerance.