What is starvation and how do you prevent it?

Starvation is when a task never gets to run because higher-priority tasks monopolize the CPU. Under preemptive priority scheduling, the highest-priority Ready task always runs, so if a high-priority task stays Ready indefinitely, because it busy-loops, never blocks, or simply has more work than the CPU can do while leaving slack, then every lower-priority task, down to the idle task, is locked out forever. The symptoms are lower-priority work that mysteriously never executes, no low-power idle, and possibly a watchdog that never gets kicked. Prevention is primarily good design: high-priority tasks should do minimal work and then block to wait for their next trigger (a semaphore from an ISR, a queue item, a delay), so they only consume CPU when there's genuinely urgent work, leaving the rest of the time for lower-priority tasks; you also ensure the overall CPU utilization is below 100% with margin so the schedule is feasible (the schedulability concern from the real-time lesson). You assign priorities by deadline urgency rather than importance-in-the-abstract, avoid long computations in high-priority tasks (break them up or move them down), and verify with run-time stats that lower-priority tasks and idle actually run. Note that classic priority scheduling has no built-in anti-starvation aging (unlike some general-purpose OS schedulers), so on an RTOS it's on you to design the workload to be schedulable.