pull down to refresh

If Phoenix had less inbound liquidity when LNproxy tried to forward the payment, it could have caused the transaction to get stuck. LNproxy might not have been able to fulfill the payment due to Phoenix’s reduced capacity, causing the HTLC to timeout and triggering the refund.
I exported the Phoenix wallet log from the event on Jan 10, 2025, UTC, to see whether it attempted to process the payments. The log indicates that it indeed tried to execute payments through MPP for amounts of 175,000 sats and 2 of 87,500 sats. But it appears that the LN proxy failed to complete these transactions, resulting in keep Phoenix purging incoming payments multiple times like image below, which ultimately led to the rollback of the transactions and the subsequent release of the HTLC by Strike. Indeed it seems you are right. Not sure if this is accurate, but this is my reasonable scenario thus far.
So, the sudden drop in your inbound liquidity (from 470k to 250k sats) could have been due to Phoenix closing and reopening channels to manage fees or balance liquidity. This isn’t unusual, but it can catch you off guard if you're not expecting it.
as mentioned by @OT, I found out that I did splice out some funds to on-chain, which explains the sudden drop in inbound liquidity.