High-voltage direct current (HVDC) has established itself as the leading technology for long-distance transmission, particularly for interconnections between countries and offshore wind farms. Sulfur hexafluoride (SF6) has traditionally been the preferred insulating medium in gas-insulated substations (GIS) due to its excellent dielectric properties; however, its high global warming potential (GWP) remains a significant drawback. Partial discharge (PD) detection serves as a critical diagnostic tool for ensuring the operational reliability of GIS systems. This study investigates the long-term PD behavior of protrusion and free metallic particle defects in HVDC GIS filled with technical air. The PD apparent charge magnitude and repetition rate evolution are analyzed using pulse sequence analysis (PSA) plots. Results indicate that PSA plots evolve and vary depending on the defect type, posing challenges for human experts and machine learning models in defect classification. Furthermore, most existing PSA plots are derived from test conditions using SF6, highlighting the need for research in alternative insulation gases such as technical air. Both conventional and unconventional PD detection methods were employed within a full-scale GIS test cell. The two defect types were subjected to voltage application for one week. The free metallic particle defect exhibited a 20% change in PD apparent charge magnitude over the test duration but showed minimal alterations in weight and physical structure. In contrast, the protrusion defect experienced a 30% increase in PD apparent charge magnitude, accompanied by significant physical changes, as revealed through microscope imaging. The observed changes in PD behavior after just one day of voltage application suggest that long-term testing in technical air is unnecessary. Similarly, PSA patterns from SF6 were successfully used to classify defects in technical air, demonstrating that knowledge transfer is possible. Finally, the similarities between the certain patterns of free metallic particles and protrusion defects in technical air highlight the need for further investigation in different test environments to refine defect classification in future studies.